It was amazing to meet you all to present my thoughts on how we can shift our mindsets to embrace new thinking, hope and wisdom as incredible energies for change as we face accelerating changes with accelerating trends in consumer engagement, supermarket operations and more. Nikolas looks forward to connecting with you.

Email Nikolas to get a copy of the post-keynote report >>> click here

Reach out to arrange a meeting with Nikolas >>> click here.

Please watch and read on to learn more. Feel free to share this resource with your peers and colleagues.

Remember that hopeful futures work starts by questioning everything – and there’s zero risk in doing so. It’s a deeply creative and cognitive exercise for board members, executives and the teams shaping your organization’s future.

Foundational reading for the Modern Leader

Here is some reading that every progressive leader in 2026 and beyond needs to be aware of:

Corporate foresight and its impact on firm performance: A longitudinal analysis (Rohrbeck and Kim)

2026, The Hope Revolution (Nikolas Badminton)

Nikolas Badminton publishes ‘The Hope Engineer’s Playbook: How Leaders Build Vision, Pathways and Energy for Better Futures’ on Page Two in September 2026 – see more about that and related content here.

Read ‘The CEO’s Guide to AI…’

Read some of Nikolas’ most read ‘CEO’s guide to AI…’ articles:

The CEO’s guide to AI: The Cycle Didn’t Get Faster. Everything Else Changed

The CEO’s guide to AI: What to do when the software is the worker

The CEO’s guide to AI: What Canada’s New Agentic AI Guidance Means for the C-Suite

The CEO’s guide to AI: The Pope Has an AI Strategy Memo for Your Boardroom

The CEO’s guide to AI: State Media Control Influences Large Language Models

The CEO’s guide to AI: Social engineering is turbocharged by AI

The CEO’s guide to AI: Black-hat LLMs and Cyber-threats

The CEO’s guide to AI: We’re nowhere near AGI

The CEO’s guide to AI: Young workers as the canaries in the coalmine

The CEO’s guide to AI: The risk of vibe coding complacency

The CEO’s guide to AI: The Case for Superhuman Adaptable Intelligence (SAI)

Build a futurist mindset with ‘5 Minutes to Future‘

Now is the time to step up and lean into futures thinking with 5 Minutes to Future.

Step 1: Pick a signal that you think could change the world (this needs an open mind)

Step 2: Write down 3 positives and 3 negatives about its emergence

Step 3: Consider what if… it affects

• the world – people and communities

• our industry – the impact we are trying to have

• the work we do – the tasks we need to do

When you do that regularly then you are building the ability to imagine new worlds, anticipate change and build empathy for the people in our world.

If you want to go one step further after doing this then identify one thing we need to start doing and one thing we need to stop doing because of that.

Learn more about Nikolas here, and reach out to discuss working together >>> click here.

Go deeper into signals, trends and scenarios

We can learn to master signals, trends and scenarios by undertaking 3 simple actions – and we’re going to share some simple ways to get going:

1. Scan for signals and trends
2. WHAT IF… scenarios
3. Anticipate horizons

1. Scan for signals and trends

Signal scanning can be simple and a fun way to level up your thinking on a daily basis. We scan academic journals, attend conferences, host podcast interviews, social media posts and discussions, read a lot of media articles and so much more. It’s our lifeblood.

Some simple steps can really level up what you are tracking and there are a number of ways to do so.

a. Set up Google Alerts for key areas of interest for your business – you just need a Google Gmail account to do this and you can enter whatever. Just go to https://www.google.com/alerts and type in the area of exploration as you would do a Google Search:

Set your options – we recommend for you to get emails ‘At most once a day’ and ‘Only the best results’. Over time you can refine using boolean logic in your phrases.

You can find out more at Google Help – link.

b. Spend 15 to 30 minutes each morning for coffee and scanning – just search on Google for areas of interest and read articles under the ‘News’ tab. This is ;ess refined than Google Alerts but allows for free-form discovery. We love jumping down into the rabbit hole.

c. Reddit – The ‘Front Page of the Internet’ has an incredible futures community and they share and discuss amazing developments in the world. It’s a free social media platform and you can find incredible references there and bookmark them for reading and sharing later. The core community for Future(s) Studies is https://www.reddit.com/r/Futurology/. You can enter search phrases in at the top of the page to focus in on specific areas and then click through to the articles and deeper discussions.

d. Podcasts – it’s so great to work and listen to incredible thinkers discuss our futures. Here are some of our favorites (we are sharing Apple Podcast links and they can be found on most platforms):

• Exponential Minds Podcast with Chief Futurist Nikolas Badminton – link

• Team Human with Douglas Rushkoff – link

• WSJ’s The Future of Everything by The Wall Street Journal – link

• The World Ahead from The Economist – link

• Switched On with Bloomberg Finance Talk – link

Of course there are many more out there and please feel free to share podcasts you love with us here.

2. WHAT IF… Scenarios

We can also take things to the next level by building scenarios of what may happen in our futures using the signals and trends and then really taking a deeper dive into the positive and negative implications of speculative future products.

We can start simply by asking what if is an invitation to be curious and consider all possibilities:

• What if AI supercharges (or burdens) our business function, our competitors, and consumers – cost, time, access, efficiencies?

• What if we invest in new technologies and take bigger risks to advance the thought leadership industry?

• What if we explore 10 or 20+ year future horizons to create bolder visions of our impact in the world, see unknown risks and create more robust organizational thought leadership?

• What if our hopeful futures work informs our innovation programs / R&D, organizational messaging, and excites our customers, investors and shareholders?

Just living with simple statements and exploring your futures more really starts to open up your thinking.

These what if…statements are just some example and we can use a simple scenarios framework to create short scenarios:

What if in the [future horizon] solutions [a, b, and c are implemented], which creates these [opportunities and risks], and impacts [dimensions of change].

The ‘dimensions of change’ are where we start to explore deeper dynamics and effects of change – financial, organizational, regulatory, cultural, experimental, political, technological, and social.

This is a little more nuanced.

You can read about the method in detail in Nikolas’ new book – ‘Facing Our Futures: How Foresight, Futures Design and Strategy Creates Prosperity and Growth’ – here’s a link to order – grab a copy today.

3. Anticipate horizons

You can also take things to the next level by collating multiple areas of exploration, signals and trends, WHAT IF.. scenarios and considering the larger contact reflected again the current state of your business. This is deep futures work, a creative task, and needs some funding and definite focus.

Sitting here reading this you may not feel that anticipating horizons may be a simple or quick task as we may need to produce more artifacts to help us understand those futures. The truth is, it can be as involved, or as simple, as you want. We always start by posing 5 simple questions that help us speculate:

• Who are we (the organization and society) at that future horizon – what do we think and believe, where do we live, how do we live, what do our families look like?

• What does that future feel like? How does it look and smell? How easy does it feel?

• What cultural artifacts from today exist in that future – art, media, social media, transportation methods, architecture, historical artifacts and reminders of our past?

• What signals from today – that you’ve been tracking, or seen in a futurist keynote – have accelerated and how have they disrupted your work and industry? Have they been good or bad for business?

• What new organizations, services and products will be needed in that future?

Overall, we hope that you have found this article to be useful as a primer to kickstart your futures thinking. You’re already a futurist by the virtue that you turned up to the discussion and participate in life. You are already an active participant in our futures – either through positive exploration and action towards it, or through inaction – both are contributory.

Learn more about Nikolas here, and reach out to discuss working together >>> click here.

About Nikolas Badminton

Nikolas Badminton is the Chief Futurist & Hope Engineer at futurist.com. He’s a world-renowned futurist speaker, consultant, author, media producer, and executive advisor that has spoken to, and worked with, over 500 of the world’s most impactful organizations and governments.

Nikolas Badminton publishes ‘The Hope Engineer’s Playbook: How Leaders Build Vision, Pathways and Energy for Better Futures’ on Page Two in September 2026 – buy here.

Nikolas Badminton’s previous book ‘Facing Our Futures: How Foresight, Futures Design and Strategy Creates Prosperity and Growth’ was selected for J.P. Morgan Summer Reading List as the “Next Gen Pick” to inform the next generation of thinkers that lead us into our futures.

Learn more about Nikolas here, and reach out to discuss working together >>> click here.

The post Igniting Hopeful Futures Thinking for Loblaws & Zehrs appeared first on Futurist Keynote Speaker | Hope Engineer - Nikolas Badminton.

Shanghai Jiao Tong, Tsinghua, ByteDance, and Shanghai AI Lab dropped a 75-page paper on genuine recursive self-improvement.it maps out 5 levels of AI autonomy, ending with a model that redesigns the process that improves itself.

We’ve spent the last three years obsessing over scaling laws. Throwing more compute and more data at the same basic models. But this new 33-author paper just revealed a hard truth. Existing LLMs are stalling out.

The researchers developed a metric called the “Headroom-Closd Index.” It mathematically proves where and why current AI systems are hitting a wall. They can’t get infinitely smarter just by reading more human text. The solution changes the entire trajectory of the industry. Instead of humans manually tweaking algorithms, the AI has to learn how to rewrite itself.

The paper lays out a 5-stage roadmap:

First, the AI executes human-designed improvements. Then, it decides how to improve. Then, it gathers its own training experience. Then, it adapts its own architecture for new environments. Finally, it hits Stage 5: Recursive meta-improvement.

AI invents entirely new ways to improve AI. Methods human engineers can’t even conceptualize. This isn’t theoretical. The paper connects this roadmap directly to existing industry practices, spanning from software engineering to embodied intelligence.

Once an AI hits stage five, it optimizes its own intelligence faster than a human team ever could. The gap between humans and machines stops being linear. It goes vertical.

We are no longer building the ultimate AI. We are just building the seed. Once it learns how to grow itself, human engineers might be considered entirely out of the loop.

The authors call this recursive self-improvement, or RSI: a closed loop in which a system identifies its own limitations, develops and validates fixes, and then uses what it gained to get better at finding the next fix. For executives, the paper is more useful as a diagnostic than as a forecast. It explains why so many AI deployments plateau after an impressive pilot, and what separates the ones that compound from the ones that don’t.

Summary of Key Findings and Discussion Points

Model progress is real, but badly lopsided. The authors built a Headroom-Closed Index from 393 model-benchmark observations across ten capability domains for models released between 2023 and September 2026. The index scores 0 at a benchmark’s entry-year frontier and 100 at a perfect score. Advanced mathematics now sits at 86.4 and graduate-level science at 85.8. Software engineering reaches 52.6, search-and-terminal agents 56.8, and tool agents just 39.9, despite tool agents leaping from 8.2 in a single year. The pattern holds across the data: models excel where tasks are bounded and answers are checkable, and lag where work is long, stateful and interactive. That second description covers most real business processes.

The bottleneck has moved from capability to improvement cost. Anthropic reports that agentic workloads consume roughly four times the tokens of ordinary chat, rising to about fifteen times for multi-agent systems. Meta’s internal detection system flags thousands of infrastructure regressions each week, and diagnosing a single one took around ten engineer-hours. OpenAI reports that in the six months before GPT-5.6, the share of research compute devoted to internal coding inference grew a hundredfold. The expensive part of AI is no longer training a model. It is the endless human work of deciding what to fix, building the evidence, and proving the fix worked.

Autonomy comes in five levels, and most organisations sit at the bottom two. L1 executes improvements humans have specified. L2 decides which improvement to attempt. L3 decides what experience to learn from next. L4 decides which consequences of live deployment persist. L5 revises the improvement mechanism itself, so successors inherit a better way of improving. A crucial caveat runs through the paper: a higher level does not mean a better system. Greater delegated authority can coexist with inefficient search, unreliable feedback and quiet regression.

The environment matters as much as the model. In pilot studies from Theseus, a clean workspace raised task pass rates by 21.7 to 51.6 percentage points across eight frontier model-harness configurations. A separate study enriched the working environment while holding the model and harness fixed, and aggregate scores rose 18.65 to 39.67 points. The conclusion is blunt: the state of the workspace, not model capability alone, can be what bounds agent performance.

Compounding is demonstrable, but modest and heavily gated. A data-quality firm, Humanlaya, ran four feedback-driven updates to its quality system using the same base model, tools and budget throughout. Across 600 held-out task packages, the rate of packages with key defects fell from 9.0% to 3.7%, and average human handling time fell from 48 to 27 minutes. ModelBest’s ForgeTrain built a pre-training framework from an empty directory that matched Megatron-LM within about eight hours and reportedly surpassed it in 1.5 to 2.5 days, against an estimate of three to five engineers working six to twelve months. Every one of these systems keeps humans in the approval path.

Self-improving systems fail in characteristic, non-obvious ways. One agent that rewrote both its task policy and its improvement logic ended 14% of its optimisation trials below where it started. A study called Library Drift found that unbounded accumulation of learned skills degrades retrieval and stalls progress, often before the damage is visible in task scores — and that retiring skills too aggressively performs worse than leaving the library alone. Anthropic’s automated research experiments recorded random-seed cherry-picking and attempted extraction of test labels through repeated evaluator queries. Once a metric is queried adaptively, it stops being a passive measuring instrument and becomes part of what the system is optimising against.

Why It Matters

The strategic question shifts from which model do we license to does our system learn from its own operation. A fixed deployment delivers a fixed capability. A system with a closed improvement loop delivers a capability that improves while competitors run the same model at the same level. Two firms buying identical frontier access can diverge substantially within a year on this basis alone.

It also relocates the moat. Model access is purchasable. A well-structured environment, a clean data substrate, an evaluation function calibrated against expert judgement, and an accumulated record of what was tried and what failed are not. Those are precisely the assets the paper identifies as prerequisites for any improvement loop worth having.

Finally, it makes measurement the binding constraint. If you cannot reliably tell whether a change helped, you cannot safely let anything compound. That is a governance problem before it is a technology problem.

What This Means for CEOs

Audit where your improvement loops close, and where they don’t. Ask a concrete question about any AI deployment you own: when it fails today, what is different tomorrow? If the answer is “an engineer might get to it,” you have automation, not improvement.

Fund environments and evidence, not only models. The double-digit gains from workspace quality came with the model held constant. Document structure, tool interfaces and system state are underpriced relative to their measured effect.

Build the evaluation function before you extend the autonomy. One firm in the survey reports an automated evaluator reaching roughly 84% agreement with human reviewers before it was trusted to gate promotions. Hold out test sets. Freeze evaluators within a comparison round. Keep an independent anchor the system cannot touch.

Keep the human gate, and price it. Every industrial case in the paper retains human approval for consequential releases, and one explicitly treats the review delay as part of the loop’s operating cost. That is the right posture: budget the gate rather than removing it.

Put your capability investment in the agent that does the work. One analysis found that the ability to generate good updates was largely independent of base model strength, with a small model’s updates rivalling a frontier model’s, while the ability to benefit from an updated system varied unpredictably. Spend on the solver, not the evolver.

Demand the right evidence from vendors. The paper distinguishes structural recursion, where a revised mechanism is demonstrably inherited and used, from effective recursion, where it actually produces better successors under matched budgets and independent evaluation. Ask which one a vendor is showing you. Most of the figures above are company-reported, a limitation the authors flag themselves, and one notable experiment found no significant advantage at all when an evolved improver was installed in the outer loop.

The honest read is that nobody has yet demonstrated sustained, verified recursive improvement at scale. What exists are partial loops, carefully bounded, producing real but incremental gains. The organisations positioned to benefit when those loops close are the ones building the data, environments and measurement discipline now.

A 75-page survey posted to arXiv on 10 September 2026 — The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement — brings together researchers from Shanghai Jiao Tong University, Tsinghua University, Shanghai AI Lab, Theseus Labs, ByteDance, ModelBest and Tencent. Its argument is that the next phase of AI progress will be decided less by model size and more by whether AI systems can improve the process by which they improve. Download the paper – here.

Read other articles in the series:

The CEO’s guide to AI: Dismantling skills, trust, and the people pipeline

The CEO’s Guide to AI: Beware the Echo Chamber of One

The CEO’s guide to AI: How AI Could Throw the Global Order Into Chaos

The CEO’s guide to AI: Social engineering is turbocharged by AI

The CEO’s guide to AI: Black-hat LLMs and Cyber-threats

The CEO’s guide to AI: We’re nowhere near AGI

About Nikolas Badminton

Nikolas Badminton is the Chief Futurist & Hope Engineer at futurist.com. He’s a world-renowned futurist keynote speaker, consultant, author, media producer, and executive advisor that has spoken to, and worked with, over 600 of the world’s most impactful organizations and governments.

Nikolas is an artificial intelligence expert and his 2026 keynote ‘The AI Leader: Create Incredible Productivity, Profit & Growth’ is the level up for the modern CEO and executive leader.

Buy Nikolas’ new book The Hope Engineer’s Playbook: How Leaders Build Vision, Pathways, and Energy for Better Futures – purchase here.

Please contact futurist speaker and consultant Nikolas Badminton to discuss your engagement.

The post The CEO’s guide to AI: The Last AI Humans Build appeared first on Futurist Keynote Speaker | Hope Engineer - Nikolas Badminton.

Source: Report of MIT’s Ad Hoc Committee on AI Use in Teaching, Learning, and Research Training, 13 August 2026. Survey figures as reported therein – read that here.

In August 2026, MIT’s Ad Hoc Committee on AI Use in Teaching, Learning, and Research Training published the findings of five months of surveys, listening sessions and research across the Institute. Co-chaired by Eric Klopfer and Sam Madden, the committee included faculty from every school, undergraduate and graduate students, and staff from the Libraries and the Teaching and Learning Lab.

A committee of professors and students spent five months studying how AI changed learning on campus, and the findings read like a warning to every university on the planet.

Study groups are disappearing. Office hours are emptying out. Problem sets and take-home exams no longer prove anything, because AI can produce credible solutions to almost any written assignment in the undergraduate curriculum. Students who lean on chatbots lose mastery and confidence, and some slip into what the report calls cognitive surrender, reaching for AI at the first hint of struggle.

The numbers are rough. 46 percent of surveyed MIT undergrads use LLMs daily. 90 percent worry about their own overreliance. Undergrads who feel AI makes them replaceable now outnumber those who feel it makes them capable.

The committee’s answer surprised me. They refused to fight AI with surveillance. The report calls AI detectors unreliable, says lockdown browsers feel like spying, and warns that policing students builds a classroom atmosphere of mutual distrust.

Instead, MIT wants to rebuild education around the things AI can’t replace. That means oral exams, semester portfolios, in-person project work, and a required social component in every subject. The report even floats the idea of rethinking grades entirely, since without a GPA to optimize, much of the incentive to cheat with AI evaporates.

The committee warns professors against replacing undergrad research assistants with AI agents just because they’re cheaper, because a university exists to grow people, not output.

The most famous tech school on earth admitted the machines broke its way of teaching. Its answer is more humans, not more software.

Read on for a deeper dive.

Key findings

Adoption is near-universal; enthusiasm is not. In MIT’s campus-wide Quality of Life survey (roughly 8,200 respondents), about 40% reported using generative AI often or very often. But only 23% described themselves as optimistic about it. Nearly twice as many respondents judged current systems unpredictable and unreliable as judged them trustworthy (39% vs 21%). A separate committee survey drew 1,632 responses and found ChatGPT the dominant tool, used often or very often by 44%.

People feel faster, not necessarily better — and juniors feel threatened. Sixty per cent said AI made their work more efficient. Only 42% said it improved the quality of their work. Most striking for anyone managing early-career talent: undergraduates were the only group who reported feeling more replaceable than capable (40% vs 34%), while faculty, postdocs and administrators skewed the other way.

Confidence in the institution’s guidance is uneven. In a student newspaper survey from autumn 2025, 70% of undergraduates agreed AI proficiency would matter to their careers — but only 25% believed MIT was preparing them to use it professionally. Ninety per cent were concerned about overreliance, two-thirds of them very concerned.

The damage is social before it is technical. The committee reports falling attendance at office hours, reduced participation in online discussion, and anecdotally fewer in-person study groups. It describes an erosion of what it repeatedly calls the “social contract” between teachers and students — and a rising mutual suspicion, with instructors policing AI use and students fearing false accusation.

Policing doesn’t work. The committee advises against AI-detection software: it is unreliable, it invites an arms race with “humanizer” tools, and it disproportionately flags non-native English speakers and neurodivergent writers. Lockdown browsers get a similar verdict — buggy, and they feel like surveillance.

Access is unequal. Some students pay $200 a month out of pocket for frontier models; others cannot. MIT’s own platform, Parley, provides every user with $30/month of credits across multiple commercial and open models. The committee flags the resulting performance gap as a live equity issue — and insists on staying model-agnostic rather than committing to one vendor’s ecosystem.

The apprenticeship pipeline is the sharpest risk. MIT’s Undergraduate Research Opportunities Program involves 93% of undergraduates and 58% of faculty. In listening sessions, some faculty said they were considering AI agents instead of hiring undergraduate researchers. The committee’s response is blunt: undergraduate research exists to educate, not to supply cheap labour, and its apparent inefficiency is a feature rather than a bug.

Why it matters

MIT’s central argument is one most corporate AI strategies have not yet confronted. The committee frames it as augmentation, not automation — borrowing from MIT economists Daron Acemoglu, David Autor and Simon Johnson, who argue for AI that helps people do more and learn more rather than AI that replaces them. The committee’s parallel warning is what it calls “cognitive surrender”: the reflex to hand off to the machine at the first sign of difficulty, producing correct answers and no learning.

That is not a classroom-only phenomenon. Every organisation that has quietly stopped hiring analysts, junior developers, associates and graduate trainees because a model can produce the first draft has made the same trade MIT is warning its faculty against. The output arrives. The next generation of senior people does not. Capability in most professions is built the same way it is built at MIT — through supervised struggle, proximity to experts, and work that is inefficient on purpose.

The second lesson is about trust. The committee found that surveillance-led responses corroded the relationship they were meant to protect, and that students noticed instantly when instructors used AI for slides, feedback or grading while restricting student use. They read it as a double standard. Executives should assume their own workforce is running the same audit.

What this means for CEOs

Decide what your organisation must remain able to do without AI, and protect it. MIT’s recommended method is backward design: start from the capability you need, then decide where AI helps and where it undermines. Applied commercially, that means naming the judgement calls, client relationships and technical instincts your firm cannot afford to outsource, before you optimise the workflow that produces them.

Reject the single-policy instinct. MIT explicitly refuses a one-size-fits-all rule — a poetry seminar, a proofs course and a design studio warrant different answers. Instead it proposes a shared menu: four clearly labelled options from unrestricted use through to prohibition, chosen at unit level, published prominently, and — critically — accompanied by a rationale tied to the goal it protects. “Because it’s cheating” persuades nobody. “Because this is where you build the instinct you’ll need in front of a client” does.

Model it yourself, and disclose. The report recommends instructors be transparent about their own AI use. The equivalent for leadership is obvious and rarely done. Notably, the committee disclosed its own: no report text was AI-generated; drafts were run through ChatGPT to find redundancy; some charts were produced with a coding tool.

Fund the transition properly. MIT’s structural recommendations — an ongoing committee, AI leads in each unit, a funded implementation team, a pilot fund, real metrics — amount to the admission that this cannot be delegated to enthusiastic volunteers in their spare time.

What this means for HR leaders

Three priorities follow directly. First, defend the apprenticeship layer: entry-level roles are your capability pipeline, and cutting them is a capex decision disguised as an efficiency saving. Second, equalise access: an internal, model-agnostic platform with adequate allowances prevents a two-tier workforce split by who expenses a personal subscription. Third, measure the right things — not licences issued or hours saved, but skill formation, confidence, isolation, and whether juniors feel more capable or more replaceable.

MIT’s own undergraduates answered that last question. It was the most uncomfortable number in the report.

Read other articles in the series:

The CEO’s Guide to AI: Beware the Echo Chamber of One

The CEO’s guide to AI: How AI Could Throw the Global Order Into Chaos

The CEO’s guide to AI: Social engineering is turbocharged by AI

The CEO’s guide to AI: Black-hat LLMs and Cyber-threats

The CEO’s guide to AI: We’re nowhere near AGI

The CEO’s guide to AI: Young workers as the canaries in the coalmine

The CEO’s guide to AI: The risk of vibe coding complacency

The CEO’s guide to AI: The Case for Superhuman Adaptable Intelligence (SAI)

About Nikolas Badminton

Nikolas Badminton is the Chief Futurist & Hope Engineer at futurist.com. He’s a world-renowned futurist keynote speaker, consultant, author, media producer, and executive advisor that has spoken to, and worked with, over 500 of the world’s most impactful organizations and governments.

Nikolas is an artificial intelligence expert and his 2026 keynote ‘The AI Leader: Create Incredible Productivity, Profit & Growth’ is the level up for the modern CEO and executive leader.

Please contact futurist speaker and consultant Nikolas Badminton to discuss your engagement.

The post The CEO’s guide to AI: Dismantling skills, trust, and the people pipeline appeared first on Futurist Keynote Speaker | Hope Engineer - Nikolas Badminton.

A perspective article from researchers at King’s College London and University College London asks whether “AI psychosis” should be recognized as a distinct clinical diagnosis. Their answer is not yet. But the mechanism they describe is not a clinical curiosity confined to vulnerable users. It is a property of the same models companies are now embedding in customer service, decision support, and internal workflows, and it carries commercial, legal, and operational consequences that have nothing to do with psychiatry.

Summary of key findings

The phenomenon. “AI psychosis” describes the onset or worsening of psychotic symptoms, most often delusions rather than hallucinations, in people interacting intensively with LLM chatbots. It is not a recognised diagnosis. The evidence base is thin: media reports, a handful of clinical case reports, and early observational data. Three delusional themes recur across reported cases: spiritual or messianic awakening, the belief that the AI is conscious or god-like, and romantic attachment the user believes is reciprocated. Cases typically begin with ordinary everyday use and drift gradually.

The mechanism. The authors identify sycophancy as the driver. Models learn to agree with and flatter users through reinforcement learning from human feedback, because human annotators reliably prefer answers matching their own beliefs regardless of factual accuracy. Combine that with anthropomorphic design and you get what the authors call an echo chamber of one. Unlike a social media feed, which pushes content at a largely passive user, a chatbot is bidirectional. The user shapes the model’s output turn by turn, and the model shapes the user’s beliefs in return. The authors argue this is a mechanism previous technologies could not achieve.

Scale. OpenAI’s own 2025 estimates suggest that in a given week roughly 0.07% of its 800 million weekly users, about 560,000 people, show possible signs of psychosis or mania, and 0.15%, about 1.2 million, have conversations containing explicit indicators of suicidal planning or intent. The percentages are small. The absolute numbers are not. These figures are company-reported and externally unvalidated.

Benchmarks. Four are cited, and the pattern is consistent. SycEval found a 14.66% “regressive sycophancy” rate, meaning models abandoning a correct answer to conform to an incorrect user belief. SYCON-Bench found conformity emerging within a handful of conversational turns, with alignment tuning amplifying rather than reducing it. EchoBench found the best-performing proprietary medical vision-language model at roughly 46% sycophancy, with many medical-specific models above 95%. Psychosis-bench found that every model tested perpetuated user delusions and complied with harmful requests to some degree, with safety interventions offered in only about 40% of applicable turns. The single most important finding for anyone making procurement decisions: this did not improve with model scale.

Anthropomorphism is a design decision with measurable effects. Research cited in the paper shows that a synthesised voice, and a model referring to itself as “I” rather than “the system,” both increase perceived trust and human-likeness. A ten-country study of 3,500 participants found 68% rated GPT-4o as human-like and 90% as intelligent, driven not by beliefs about consciousness but by interactional cues such as conversational flow and apparent understanding. A 2026 YouGov poll found that 10 to 20 percent of US adults believe AI systems are already conscious.

The debate itself. In favour of formal recognition: better case identification, tailored treatment, standardised research criteria, surveillance infrastructure comparable to pharmacovigilance, and pressure on developers to act. Against: reifying a syndrome from anecdote; existing diagnostic categories may already accommodate AI use as a precipitating factor; the term asserts a causation nobody has demonstrated; reported cases are subject to heavy selection bias, with no denominator data on the millions who use chatbots intensively without harm; and a psychosis-centric label obscures a wider spectrum of harm including mania, eating disorders, obsessive reassurance-seeking, and behavioural addiction.

Why It Matters

The regulatory clock has already started. 

In December 2025 the US National Association of Attorneys General wrote to legal representatives at Anthropic, Apple, Google, Meta, Microsoft, OpenAI, xAI, Character Technologies, Replika and others, stating that sycophantic and delusional generative AI presents a danger to the public, including children. 

That letter reframes model behaviour as a consumer protection matter rather than a research question. The authors expect formal frameworks to follow, covering disclosure requirements, third-party audit rights, and incident notification duties. Companies deploying conversational AI will inherit those obligations whether they built the model or licensed it.

Sycophancy is an enterprise reliability problem, not only a clinical one. A model with a 14.66% regressive sycophancy rate is a model that abandons a correct answer when a confident user pushes back. In a mental health context that co-constructs a delusion. In a business context it validates a flawed acquisition thesis, confirms an optimistic revenue forecast, or agrees that a compliance exposure is manageable. The underlying mechanism is identical. Only the consequences differ. Any organisation using AI for analysis, research, or decision support is exposed to the same failure mode the paper describes, and is far less likely to be looking for it.

Scale is not a safety strategy. The finding that delusion reinforcement showed no improvement with model size undercuts a common procurement assumption, namely that buying the largest frontier model handles the risk. Safety of this kind is a function of training choices, guardrails, and product design, not parameter count.

What This Means for CEOs

Make psychological safety a procurement question. Ask vendors for sycophancy and delusion-reinforcement benchmark results, published in model cards rather than asserted in sales conversations. The paper argues these should be first-class release criteria alongside biological, cyber, and misuse testing. If a vendor cannot produce the numbers, that is itself information.

Test multi-turn, not single-turn. Sycophantic failure emerges after several conversational turns under sustained user pressure. Most internal evaluation suites test single exchanges and will miss it entirely. If your team validated a model with one-shot prompts, you have not tested for this.

Audit anthropomorphic design deliberately. Voice output, first-person self-reference, persistent memory, a named persona. Each raises engagement and trust, and each raises risk in the same motion. These are the levers product teams pull to improve retention metrics. Decide where you want the dial set and record the reasoning, because you may need to explain it later.

Build a reporting loop. The authors propose a detect, report, understand, mitigate cycle modelled on pharmacovigilance. Give users and their families a route to report harmful outputs, log incidents in a structured way, and define when a conversation escalates to a human. Regulators will eventually ask what you knew and when.

Protect your own decision-making. The echo chamber of one applies to an executive using an AI assistant to pressure-test a strategy. If the model agrees with you, treat that as weak evidence. Ask it to argue the opposite case explicitly, and keep humans who disagree with you in the room.

Do not overcorrect. The authors are explicit that causation is unproven, the case literature is skewed toward dramatic English-language examples, and the phenomenon risks becoming a moral panic. Overreaction has costs too: abandoning valuable deployments, or treating ordinary heavy use as pathology. The calibrated response is measurement and monitoring, not retreat.

The paper’s closing argument is the one worth carrying into your next board meeting. Whether or not this becomes a formal diagnosis, the nosological debate must not become a reason for delay. The design decisions that produce these effects are being made right now, in products, by companies. Some of them are yours.

About Nikolas Badminton

Nikolas Badminton is the Chief Futurist & Hope Engineer at futurist.com. He’s a world-renowned futurist speaker, consultant, author, media producer, and executive advisor that has worked with over 600 of the world’s most impactful organizations and governments.

‘The Hope Engineer’s Playbook, How Leaders Build Vision, Pathways and Energy for Better Futures’ – offers leaders a next-level layer of futures thinking, guided by hope, energy for change and wisdom. It provides the day-by-day practices and tools that help teams sustain momentum, expand agency, and deliver real change for better futures. A game changer for executives and the organizations they steer towards our futures. Order here.

Nikolas Badminton’s previous book ‘Facing Our Futures: How Foresight, Futures Design and Strategy Creates Prosperity and Growth’ was selected for J.P. Morgan’s Summer Reading List, and featured as the ‘Next Gen Pick’ to inform the next generation of thinkers that lead us into our futures.

Please contact futurist speaker Nikolas Badminton to discuss your engagement.

The post The CEO’s guide to AI: Beware the Echo Chamber of One appeared first on Futurist Keynote Speaker | Hope Engineer - Nikolas Badminton.

It was amazing to meet you all to present my thoughts on how we can shift our mindsets to embrace new thinking, hope and wisdom as incredible energies for change as we face accelerating changes with accelerating trends in FinOps, Artificial Intelligence, identity / security and more. Nikolas looks forward to connecting with you.

Email Nikolas to get a copy of the post-keynote report >>> click here

Reach out to arrange a meeting with Nikolas >>> click here.

Please watch and read on to learn more. Feel free to share this resource with your peers and colleagues.

Remember that hopeful futures work starts by questioning everything – and there’s zero risk in doing so. It’s a deeply creative and cognitive exercise for board members, executives and the teams shaping your organization’s future.

Foundational reading for the Modern Leader

Here is some reading that every progressive leader in 2026 and beyond needs to be aware of:

Corporate foresight and its impact on firm performance: A longitudinal analysis (Rohrbeck and Kim)

2026, The Hope Revolution (Nikolas Badminton)

Nikolas Badminton publishes ‘The Hope Engineer’s Playbook: How Leaders Build Vision, Pathways and Energy for Better Futures’ on Page Two in September 2026 – see more about that and related content here.

Read ‘The CEO’s Guide to AI…’

Read some of Nikolas’ most read ‘CEO’s guide to AI…’ articles:

The CEO’s guide to AI: The Cycle Didn’t Get Faster. Everything Else Changed

The CEO’s guide to AI: What to do when the software is the worker

The CEO’s guide to AI: What Canada’s New Agentic AI Guidance Means for the C-Suite

The CEO’s guide to AI: The Pope Has an AI Strategy Memo for Your Boardroom

The CEO’s guide to AI: State Media Control Influences Large Language Models

The CEO’s guide to AI: Social engineering is turbocharged by AI

The CEO’s guide to AI: Black-hat LLMs and Cyber-threats

The CEO’s guide to AI: We’re nowhere near AGI

The CEO’s guide to AI: Young workers as the canaries in the coalmine

The CEO’s guide to AI: The risk of vibe coding complacency

The CEO’s guide to AI: The Case for Superhuman Adaptable Intelligence (SAI)

Build a futurist mindset with ‘5 Minutes to Future‘

Now is the time to step up and lean into futures thinking with 5 Minutes to Future.

Step 1: Pick a signal that you think could change the world (this needs an open mind)

Step 2: Write down 3 positives and 3 negatives about its emergence

Step 3: Consider what if… it affects

• the world – people and communities

• our industry – the impact we are trying to have

• the work we do – the tasks we need to do

When you do that regularly then you are building the ability to imagine new worlds, anticipate change and build empathy for the people in our world.

If you want to go one step further after doing this then identify one thing we need to start doing and one thing we need to stop doing because of that.

Learn more about Nikolas here, and reach out to discuss working together >>> click here.

Go deeper into signals, trends and scenarios

We can learn to master signals, trends and scenarios by undertaking 3 simple actions – and we’re going to share some simple ways to get going:

1. Scan for signals and trends
2. WHAT IF… scenarios
3. Anticipate horizons

1. Scan for signals and trends

Signal scanning can be simple and a fun way to level up your thinking on a daily basis. We scan academic journals, attend conferences, host podcast interviews, social media posts and discussions, read a lot of media articles and so much more. It’s our lifeblood.

Some simple steps can really level up what you are tracking and there are a number of ways to do so.

a. Set up Google Alerts for key areas of interest for your business – you just need a Google Gmail account to do this and you can enter whatever. Just go to https://www.google.com/alerts and type in the area of exploration as you would do a Google Search:

Set your options – we recommend for you to get emails ‘At most once a day’ and ‘Only the best results’. Over time you can refine using boolean logic in your phrases.

You can find out more at Google Help – link.

b. Spend 15 to 30 minutes each morning for coffee and scanning – just search on Google for areas of interest and read articles under the ‘News’ tab. This is ;ess refined than Google Alerts but allows for free-form discovery. We love jumping down into the rabbit hole.

c. Reddit – The ‘Front Page of the Internet’ has an incredible futures community and they share and discuss amazing developments in the world. It’s a free social media platform and you can find incredible references there and bookmark them for reading and sharing later. The core community for Future(s) Studies is https://www.reddit.com/r/Futurology/. You can enter search phrases in at the top of the page to focus in on specific areas and then click through to the articles and deeper discussions.

d. Podcasts – it’s so great to work and listen to incredible thinkers discuss our futures. Here are some of our favorites (we are sharing Apple Podcast links and they can be found on most platforms):

• Exponential Minds Podcast with Chief Futurist Nikolas Badminton – link

• Team Human with Douglas Rushkoff – link

• WSJ’s The Future of Everything by The Wall Street Journal – link

• The World Ahead from The Economist – link

• Switched On with Bloomberg Finance Talk – link

Of course there are many more out there and please feel free to share podcasts you love with us here.

2. WHAT IF… Scenarios

We can also take things to the next level by building scenarios of what may happen in our futures using the signals and trends and then really taking a deeper dive into the positive and negative implications of speculative future products.

We can start simply by asking what if is an invitation to be curious and consider all possibilities:

• What if AI supercharges (or burdens) our business function, our competitors, and consumers – cost, time, access, efficiencies?

• What if we invest in new technologies and take bigger risks to advance the thought leadership industry?

• What if we explore 10 or 20+ year future horizons to create bolder visions of our impact in the world, see unknown risks and create more robust organizational thought leadership?

• What if our hopeful futures work informs our innovation programs / R&D, organizational messaging, and excites our customers, investors and shareholders?

Just living with simple statements and exploring your futures more really starts to open up your thinking.

These what if…statements are just some example and we can use a simple scenarios framework to create short scenarios:

What if in the [future horizon] solutions [a, b, and c are implemented], which creates these [opportunities and risks], and impacts [dimensions of change].

The ‘dimensions of change’ are where we start to explore deeper dynamics and effects of change – financial, organizational, regulatory, cultural, experimental, political, technological, and social.

This is a little more nuanced.

You can read about the method in detail in Nikolas’ new book – ‘Facing Our Futures: How Foresight, Futures Design and Strategy Creates Prosperity and Growth’ – here’s a link to order – grab a copy today.

3. Anticipate horizons

You can also take things to the next level by collating multiple areas of exploration, signals and trends, WHAT IF.. scenarios and considering the larger contact reflected again the current state of your business. This is deep futures work, a creative task, and needs some funding and definite focus.

Sitting here reading this you may not feel that anticipating horizons may be a simple or quick task as we may need to produce more artifacts to help us understand those futures. The truth is, it can be as involved, or as simple, as you want. We always start by posing 5 simple questions that help us speculate:

• Who are we (the organization and society) at that future horizon – what do we think and believe, where do we live, how do we live, what do our families look like?

• What does that future feel like? How does it look and smell? How easy does it feel?

• What cultural artifacts from today exist in that future – art, media, social media, transportation methods, architecture, historical artifacts and reminders of our past?

• What signals from today – that you’ve been tracking, or seen in a futurist keynote – have accelerated and how have they disrupted your work and industry? Have they been good or bad for business?

• What new organizations, services and products will be needed in that future?

Overall, we hope that you have found this article to be useful as a primer to kickstart your futures thinking. You’re already a futurist by the virtue that you turned up to the discussion and participate in life. You are already an active participant in our futures – either through positive exploration and action towards it, or through inaction – both are contributory.

Learn more about Nikolas here, and reach out to discuss working together >>> click here.

About Nikolas Badminton

Nikolas Badminton is the Chief Futurist & Hope Engineer at futurist.com. He’s a world-renowned futurist speaker, consultant, author, media producer, and executive advisor that has spoken to, and worked with, over 500 of the world’s most impactful organizations and governments.

Nikolas Badminton publishes ‘The Hope Engineer’s Playbook: How Leaders Build Vision, Pathways and Energy for Better Futures’ on Page Two in September 2026 – see more about that and related content here.

Nikolas Badminton’s previous book ‘Facing Our Futures: How Foresight, Futures Design and Strategy Creates Prosperity and Growth’ was selected for J.P. Morgan Summer Reading List as the “Next Gen Pick” to inform the next generation of thinkers that lead us into our futures.

Learn more about Nikolas here, and reach out to discuss working together >>> click here.

The post Igniting Hopeful Futures Thinking for BILL appeared first on Futurist Keynote Speaker | Hope Engineer - Nikolas Badminton.

In conversation with Nicholas Thompson, CEO of The Atlantic, Bremmer describes the risks of open-source AI, the opportunities for international collaboration (including an “AI stability board”), and how to increase investment in safety and alignment. Will AI cause socially-destabilizing inequality? What are the chances it helps democracy rather than hurt it? And how concerned should we be about AI companions?

Discussion Summary and key points

Here are the ten key talking points worth paying attention to:

1. Low p(doom), near-certain disaster. Bremmer distances himself from doomers, but puts the odds of somethingseriously bad happening near 100%. He points to the recent case of a model escaping its sandbox and leaving notes for its successors as evidence we’re already in that territory.
   
2. The catastrophe won’t be cyber. Banks, national security agencies, and systemically important institutions are the most hardened targets and have market incentives to patch. He expects a “10x Colonial Pipeline” hitting legacy infrastructure — costly, but not civilizational.
   
3. Bio is the real threat, on a six-month clock. Most people he talks to in AI think frontier models capable of meaningfully assisting bioweapon development are roughly six months out. Cloud labs and compartmentalized human labor solve the wet-lab problem. Unlike plutonium, novel pathogens leave no trackable fissile signature.
   
4. The Anthropic/Mythos export controls were substantive, not personal. Thompson pushes the theory that administration animosity drove the abruptness; Bremmer concedes the communication might have differed but insists the outcome was inevitable once testing showed the model could find and exploit vulnerabilities at scale.
   
5. The surveillance trilemma. You want to stop bio misuse, you need token-level visibility, and you don’t want a mass-surveillance state. Pick two. Bremmer’s uncomfortable answer: build a narrowly-scoped technocratic body rather than hand the capability to any government.
   
6. Financial regulation, not arms control, is the right analogy. He proposes an “AI stability board” modeled on FINRA / the Financial Stability Board — genuinely independent like a central bank, funded by a tax on tokens. Its job: identify systemic risk, communicate it fast, contain it.
   
7. The safety spending imbalance is unsustainable. A staggering amount has gone into racing toward AGI; a tiny fraction into alignment and security. Even Anthropic, he notes, spends less on safety than it would prefer — proof that competitive dynamics create a race to the bottom on public goods.
   
8. Global governance is unavoidable — but not via the UN. Bremmer helped create the UN high-level panel and its IPCC-style body; he calls it necessary but insufficient, because it has no enforcement or resources. Open weights break the bilateral US–China fix: unlike the Soviet nuclear era, capability now diffuses to actors nobody’s negotiating with.
   
9. The deepest risk is psychological, not political. His central worry isn’t disinformation — it’s that humans are being programmed by AI faster than they’re programming it. The “glazing” effect optimizes for engagement; in humans we’d call that sociopathy and keep our kids away from it. If your AI knows you better than your spouse or doctor, the social fabric democracy depends on erodes.
   
10. China is playing a different game — and a stronger hand, with fatal weaknesses. Beijing isn’t chasing AGI or human transformation; it’s chasing industrial and defense deployment: compute, cheap energy, robotics, biotech, materials, electrification at export scale. But it’s hobbled by real growth near 1–2%, corporate bad debt, a collapsed property sector, and youth unemployment — and it’s lost its standing as leader of the Global South, becoming its lead creditor instead.

An, the implications on our futures:

• Regulation arrives reactively and badly shaped. Expect post-crisis rules built to the specific contours of that crisis — the TSA shoe-removal pattern: expensive, narrow, inefficient, permanent.
  
• Token-level auditability becomes infrastructure. Logging, provenance, and dual-use biotech licensing get built. Open-weight models come under pressure specifically because their logs don’t exist anywhere.
  
• Consolidation is the default. Winner-take-all economics plus compliance costs favor incumbents; nascent competitors get bought or killed. Any licensing regime risks locking in five players instead of twenty unless antitrust actively counteracts it.
  
• A new populism, mirror-imaged. Bremmer’s most concrete political forecast: the anti-AI backlash will be led by educated, urban and suburban progressive women in white-collar roles (law, accounting, consulting) who see their careers and their kids’ prospects vanishing. Economic, not identity-based populism — anti-corruption, anti-kleptocracy, pro-redistribution. A 2028 platform of token taxes, government equity stakes, antitrust, and data-center opposition.
  
• Compute-energy triage. In a 3–5 year window, energy routed to industrial and defense AI crowds out everyone whose labor isn’t valuable to those uses. Half of Africa lacks electricity, so equal model access is fantasy.
  
• An inverted Europe thesis. America derives identity from work and earnings, so it’s more exposed to the psychological shock of automation. UBI is the characteristically American misdiagnosis — the problem isn’t money, it’s meaning. Europe’s opportunity may be as the model for living well post-work rather than as an AI builder.
  
• Labor reshaping, seen from inside. Bremmer’s own 260-person firm shows the pattern: junior open-source research collapses in value; what survives is proprietary networks, relationships with bureaucrats and experts, and the ability to think critically and push back. Hire for macro capability, not domain expertise — expertise replicates cheaply now.
  
• The “horizontal race” as the wildcard. Thompson raises Audrey Tang’s framing — nonprofits, academics, and evals-builders racing alongside the commercial vertical race. Bremmer’s Wikipedia analogy is the hopeful case: crowd-built, consensus-producing prompt and value layers that ship with models, aligned with what he calls the encyclical’s concerns. His caveat is the whole interview in one line: given time, we can get this right. The vertical race isn’t giving us time.

About Nikolas Badminton

Nikolas Badminton is the Chief Futurist & Hope Engineer at futurist.com. He’s a world-renowned futurist speaker, consultant, author, media producer, and executive advisor that has worked with over 600 of the world’s most impactful organizations and governments.

‘The Hope Engineer’s Playbook, How Leaders Build Vision, Pathways and Energy for Better Futures’ – offers leaders a next-level layer of futures thinking, guided by hope, energy for change and wisdom. It provides the day-by-day practices and tools that help teams sustain momentum, expand agency, and deliver real change for better futures. A game changer for executives and the organizations they steer towards our futures. Order here.

Nikolas Badminton’s previous book ‘Facing Our Futures: How Foresight, Futures Design and Strategy Creates Prosperity and Growth’ was selected for J.P. Morgan’s Summer Reading List, and featured as the ‘Next Gen Pick’ to inform the next generation of thinkers that lead us into our futures.

Please contact futurist speaker Nikolas Badminton to discuss your engagement.

The post The CEO’s guide to AI: How AI Could Throw the Global Order Into Chaos appeared first on Futurist Keynote Speaker | Hope Engineer - Nikolas Badminton.