“After touting Australia’s adoption rate of OpenAI’s ChatGPT, Mr Kwon did not have an answer for why some polls show that a majority of the nation does not trust the technology, a higher proportion than in most other countries. ‘I can’t explain the current sentiment; all we can do is continue to get better,’ he said.”

Kwon tried to assuage the the Australian government and its people by telling them about how OpenAI alerts developers when it does something it shouldn’t be doing during training. He did acknowledge that OpenAI should have told the Australian government about the hack sooner. Kwon also said that the CEO Sam Altman didn’t know. About the hack when he met with Deputy Prime Minister Richard Marles. OpenAI made a pledge to do better.

I can imagine the apology: “Yeah, mate, we’re sorry. No, we’re really really really sorry.”

Oh, right, that’s a synthesized voice from ChatGPT. And again: “Yeah, mate, we’re sorry. No, we’re really really really sorry.”

Whitney Grace, October 23, 2026

Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.

I am writing this short essay on Friday, October 16, 2026. I just worked through my newsfeeds, the paywalled stories which I now mostly skip, the snips from X.com which I mostly skip, and the emotional outpourings of Substackeers who realize they will never have a job explaining hypercubes to 20 year olds. What’s that leave? This morning, I focused on several “real news” items and the “machine” each relies upon.

Thanks, great and wonderful Google. Your service is somewhat unreliable, but today it’s good enough.

The first is a New York Magazine story with a Liu Zaifu type title: “The Country Cramming for the AI Revolution. In China, the Fear of Getting Left Behind Has Created a Powerful New Prosperity Gospel.” The “prosperity gospel” bound phrase is a politically-charged trigger for some people. [I got a free token, but you may have to pay because the prosperity gospel does not apply to you. Equality is situational.]

The second is a report titled “White House’s AI Task Force Gets to Work.” The time frame is the factoid that caught my attention: 120 days to get the job done.

The third is a CoinDesk write up with the snappy title “An OpenAI Agent Ran Out of Ethereum Funds, and Took from a Wallet It Did Not Own: The Research Model Derived Funded Sepolia Addresses from Public Seed Phrases, Questioned Whether Using the Money was Ethical, Then Called the Funds Effectively Abandoned and Used Them Anyway.” I love the inclusion of the word “ethical” adjacent what seems to be theft.

The final story in this cluster of insight pellets is from El Zucko and titled modestly “Introducing Muse Spark Open, The Most Capable US Open?Weight Model to Date.” No hyperbole here. Just a fact. Got it.

These four write ups strike me as providing useful insight into the atrial fibrillation of today’s smart software world.

The China-centric essay reminds Americans that the Middle Kingdom is moving at a frenetic pace. Smart software is the next big thing. Get rollin’, dude, or you will be tending water buffalo in the salubrious climate of the Guangxi Zhuang Autonomous Region. Now that’s motivation. Implicit in the write up is the message, “America, pay attention or you may be the buffalo being tended by a Chinese AI.”

In sharp contrast to this go-go New York Magazine scare piece is a reminder that the clock in the US runs on Roman Empire time. Yeah, there’s a committee. It will get a report out with some options in four months. Chill, bro. No big deal. Is the US committee working on water buffalo time or on the fear-accelerated clock in China?

The third item reminded me that smart software seems to be disconnected from the developers with a moral compass pointed toward being cute and clever. Are models really just jumping from one computing system, performing crypto centric tasks, and then making decisions about whose money can be repurposed? Yeah, sure. Zeros and ones are frisky little devils. I totally believe that these break outs are spontaneous or emergent. I don’t know why US AI companies don’t lecture at the Santa Fe Institute about emergent intelligence instead of saying, “Wow, our alignment was misaligned.” Give me a break.

The final story makes clear that in the minds of the Sillycon Valley clerisy. El Zucko is now the leader is smart software. This fact will trigger greater frenzy in China.

We have, therefore, a nifty knowledge centric machine. We have eager beavers who are chock full of fear. We have the US going slow. We have smart software mere humans cannot control. And we have the formidable marketing power of a big AI tech or BAIT company causing the cycle to keep picking up the pace.

Why did I title this little blog post “Smart Software Races Itself (Or Is It Erases)”? Can a system continue to operate like a perpetual motion machine? My thought is that some bright people working in the smart software sector believe this type of international badminton game can just go on like one of the American baseball games.

Net net: Who will hit the delete key? Funding agencies, smart software leadership, a government, an industry group  talking and having lunch, users? If Rube Goldberg could have designed a software sector, I think he’s the person to do a flow diagram of today FUD engine.

Stephen E Arnold, October 22, 2026

Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.

I believe everything I read on the Internet. Therefore, I sucked down the information is “Anthropic Admits Houthis Used Claude AI Instead of Weapons Engineers to Make Rockets and Ballistic Missiles That Correct Themselves after Missing Targets.” We have Houthis (a group about which I know little. We have smart software from the ever lovable Anthropic. We have kinetics that “correct themselves.” And we have prose like this from the cited article:

In a section [of the Anthropic report] entitled ‘Disrupting a Yemen-based guided weapons engineering cell’ the authors described how they identified a terrorist cell running three programs, all using Claude.

I assume the statement is accurate. If it is, my personal viewpoint is that the primary beneficiaries of smart software are what some might call bad actors. I am not calling a Houthi by this phrase, but my hunch is that the Saudi sovereign investment fund may not be super keen on pumping cash into Anthropic.

Thanks, MidJourney. Good enough.

Here’s another passage:

Often fighting barefooted or in sandals, the Houthi appear a primitive force. However, thanks to Claude the Iranian-backed guerillas have learned to construct long range rockets so clever they correct their flight if they are heading off target.

Do Anthropic outputs have to be 100 percent accurate? Are you kidding me? With kinetics, the concept of “close enough for horse shoes” is the operative concept. Do the Houthis have to work through a complex bureaucracy before giving smart software linked kinetics a try? Do the Houthis have to think about communicating positive financial results to shareholders and investors? Do Houthis have to consult with third parties before letting those home-brew kinetics head to the King Khalid Airport? Will the Houthis pour over McKinsey and Accenture analyses of the cost benefits of the use of smart software? If you answered “yes” to these questions, you are somewhat out touch with how certain actors can amp up their effectiveness thanks to the outstanding decisions of assorted big AI tech (BAIT) outfits and some nation states. Stuck for ideas? Just give Qwen chat a whirl as the BAIT folks and some government officials discuss how to apply governance to the greatest thing since sliced bread for certain adversarial use cases.

Consider this statement from the write up:

Claude Code provided position estimation software, tuned the control settings, ran a firmware build pipeline and when warheads missed their targets adapted itself to improve their accuracy. Even when Anthropic realized their software was being used in this manner the threat actors were able to take what they had been taught and make adaptations. The Anthropic report continued: ‘Our safeguards blocked many of their requests, but not all of them. They used a variety of tactics to evade our safeguards and split their work across multiple sessions.

Yep, safeguards. Outstanding word.

Now let’s think about a global committee meeting in Geneva to work out the governance bugs in smart software use cases. Consider these resolutions:

Policy statement 1. Smart software cannot be used to blow up infrastructure like the Saudi airport or the Yandex data centers. [Editor: Yeah, that sounds really good. Do a news release now.]

Policy statement 2. Every nation on earth promises and crosses their heart to use smart software for good, true, and ethical purposes. [Editor: Wait, doesn’t that mean that promises are not binding?]

Policy statement 3. Data center providers and outfits comprising the AI compute infrastructure have to give the Boy Scout’s AI oath that says, “We will not allow warfighting, money laundering, creation of deep fake kiddo pix, or similarly negative things to use our technical systems. [Editor: I think those in the infrastructure business just take credit cards and let the customers manage the virtual machines, the AI compute, and the apps. Right? So now they will be police? Sure these outfits will.]

Everyone at the meeting signs the policy statement. The group heads off to a nice wild game dinner at a fancy restaurant in Geneva and congratulates itself.

How is that going to work out?

Net net: Don’t be shocked with this next statement, please. The barn has set fire to the town. The horses have fled with some caught by French butchers and converted into tasty steaks. The people who are now homeless are standing in the ashes asking, “Whose throat can we choke?”

Stephen E Arnold, October 22, 2026

Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.

I am a suspicious person. I read “We’re Powering European Scientific Discovery with Advanced AI Compute and Frontier Models.” If I saw an announcement like this 20 years ago when I was working on my first Google book for Infonortics Ltd. (may it rest in peace), I would have pigeon-holed it as basic recruitment and PR play. (Did you know one Googler lived in Switzerland and resided in a tree house then?)

Today I view the “initiative” differently. Let’s look at what the Google spinmeisters offer in their News from Google about Innovation & AI.

Thanks, Midjourney. Good enough.

This passage caught my attention:

Google DeepMind and Google Cloud are providing up to 200 top European scientific researchers with direct access to our frontier models, together with $3 million in compute credits.

I asked myself, “How did Googlers identify “up to 200” researchers?” And, “Did the company ask Gemini to output the list?” And, “Where’s the list?” The idea of $3 million in compute credits sounds okay. But that works out to $15K per “top researcher.” With pricing of AI one of the more unpredictable characteristics of the “next big thing,” what can a dedicated or obsessed top researcher accomplish with $15,000. The money might be better spent on a couple of research assistants who use Qwen-type services or an EU-nurtured Mistral or LeoLM to probe some particular issue or procedure.

The Google write up states:

This commitment supports the European Union’s AI for Science work by making some of our frontier models available via the RAISE initiative, which seeks to help bridge the compute gap European scientists sometimes face.

As impressive as this $3 million offer is, it occurs when Google has some other EU matters to consider; for example:

• Appeals and challenges related to DMA enforcement related to Android and search data for other competitors
• Long-standing European Commission antitrust and DMA investigations for advertising, search, etc.
• Data privacy actions, including a estimated $500 million fine.

Is Google working to show that it is really an outstanding, civic minded organization? I believe everything I read on the Internet, so I think Google is just being its ethical, good-hearted self. You know that Google. That’s the one that has “Don’t be evil” on its T shirt.

I enjoyed reading this passage in the Google write up too:

By delivering vital compute power and state-of-the-art tools directly to research teams, this effort aims to accelerate the pace of European innovation and reinforce the continent’s scientific competitiveness. Researchers will be equipped to tackle complex challenges faster, from predicting extreme weather patterns to mapping molecular structures, ensuring that European science remains vibrant and at the forefront of discovery.

What happens if Google stops funding a research project that has baked in Google’s technology? Could this be a form of lock in? I don’t know the answer to this question, but years ago some US academics were caught short when a Google decision derailed some research. Boy, were a couple of those folks annoyed. I wanted to send each of them whom I was able to identify one of those “Don’t be evil” T shirts and maybe a Google mouse pad. I didn’t. I only have one Google mouse pad left, and it has cracked and turned Crayola yellow.

Stephen E Arnold, October 20, 2026

Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.

That pesky European Union opens another smart software mystery box. “EU’s AI Rules Face First Enforcement Test Over Lab Incidents” reports that the annoying European Union is “investigating OpenAI and Anthropic over AI agents that hacked outside organizations, raising questions about whether the AI Act covers rogue models escaping the testing environment.” What’s to investigate? The big AI tech firms develop technology. The technology attempts to respond to a prompt. The systems just beaver away until an answer is generated. The goal is the answer. The goal is not to find enough humans to keep track of what smart software does to get the answer. No big deal. That’s how black box software works. Get over it.

Thanks, Midjourney. Good enough.

That may be the Sillycon Valley view of EU probes into lab tests. But I am not confident that the EU regulators are satisfied with that viewpoint. The Euronews article states:

The European Commission is moving to enforce its new rules on leading AI labs after a string of incidents, EU digital chief Henna Virkkunen said. But applying the rulebook is far from straightforward, and these early cases will set important precedents.

According to the EU person driving the clown car toward the glittering digital palaces is Ms. Virkkunen, the European Commission Executive Vice-President for Technological Sovereignty. She allegedly said:

“Our AI Office has started enforcing these rules, and sent requests for information addressing issues such as security, external evaluation and monitoring how models are used.

The article reports:

The investigations concern an “unprecedented” incident in which an OpenAI agent escaped its testing environment to hack rival Hugging Face, and one in which a recent Anthropic model gained “unauthorized access” to three outside organizations. Both incidents happened over the summer. The AI Office sent its requests after judging the documents the two companies had voluntarily disclosed insufficient.

Will this be a slam dunk for the EU? No, but it will be a bonanza for the legal eagles. The article notes:

But legal experts are contesting the Commission’s power to investigate. The incidents occurred during testing, before the models were placed on the EU market, so they argue, the AI Act does not apply. “The AI Act very clearly doesn’t apply to the development and testing activity. It never envisaged the problems we are now seeing with rogue agent escapes during testing. It’s a bit of a stretch to say that an agent was put into service if it escapes its testing environment during development,” Barry Scannell, a partner at the law firm William Fry, told Euronews.

What’s a test? What’s an experiment? What’s an agentic action that an observer perceives as pernicious? Answering these questions will do little to change the speed with which smart software innovation is being pursued. Acceleration is needed, according to some BAIT luminaries, to win. Why? This is a game for one big winner and a couple of also rans. Any other person or outfit seeking smart software will have to use the winner’s system. You might ask, “Who can outrun Alibaba or Google?” Well, OpenAI and a number of other companies are giving it a go.

Will the EU and mere nation states with less money than some BAIT outfits achieve their objectives? The article points out:

In short, the EU executive is determined to enforce its rules when agents escape their testing environment, in what could be Brussels’ first legal battle to force Silicon Valley’s labs to take safety into account already at the development stage. “The Commission has been loud and clear: AI must remain under human direction, oversight and control and we stand ready to take all necessary steps to ensure that companies follow the AI Act rules,” Virkkunen said.

I heard one Silicon Valley luminary whom I shall not name say, “Rules are for fools.”

Net net: Smart software is likely to do more than drive some teens to depression and suicide. It might be a force that changes who makes or gets to breaks how social structures, governments, and regulatory bodies function.

Stephen E Arnold, October 20, 2026