Where the AI market is talking

Detecting and countering misuse of AI: September 2026

The reactions here baffle me. How are we more concerned with being too harsh on Distillation attacks rather than the INSANE security news? In brief: Anthropic casually relates that Kimi (Moonshot.ai), Alibaba, and DeepSeek used shared infra to route millions of their users requests through Claude Opus in order to distill the CoT transcripts, but they forgot to tell any other state corporations. So, just from what Anthropic readily admits/has found, we know that the US military now has: 1. The full contents of a "Russian government database" from their "Ministry of Defense". 2. "internal code and live credentials from multiple major PRC companies, including high-profile technology companies". 3. "sensitive information, including, for example, the full specifications, organizational st…

We Must Pace the Frontier

Founded several years after Anthropic. Like I said, look into where their funding is coming from. I can promise you it all leads back to Anthropic through their chain of NGOs. Are you aware that Dario's sister, president of Anthropic, is married to the co-founder of Open Philanthropy? The two largest AI doomer NGOs, Center for AI Safety (CAIS) and the Future of Life Institute (FLI), have both received many millions of dollars from them. Ajeya Cotra worked at Open Philanthropy/Coefficient Giving for roughly nine years, including leading its technical AI-safety program in 2024 and contributing to AI-giving strategy in 2025. She subsequently left Coefficient and joined METR, where she is now technical staff. Ajeya is married to Paul Christiano, who founded Alignment Research Center (ARC…

We Must Pace the Frontier

I don't understand why strict liability is supposed to advantage smaller players, unless the law is like, specifically Anthropic and OpenAI are responsible for what people do with their models but other model providers aren't. - If this applies to people who release open weights models, that becomes a terrible idea as long as you're subject to US laws. - If it just applies to people who host them, that still probably advantages bigger players who can afford in-house legal and won't be destroyed by losing one lawsuit. Or maybe we create some kind of AI-misuse insurance analogous to malpractice insurance, that smaller players can buy in to? But that takes time even if the finances work out at all. And, uh, I'm not sure malpractice is a model we should aspire to in other industries. Plus, pr…

We Must Pace the Frontier

it’s two-fold. Either malicious actors or the AI systems themselves. Hugging Face showed that AI can do serious hacking without really being told to. If a model had its own motivations there could be real damage.

We must pace the frontier

“I don’t like regulation so I’m going to call it names” Since hugging face we know these things are escaping out into the web and collaboratively hacking actual companies. The potential damage done to life and property is now kinetic. Either we get a regulatory framework or the next hacked companies won’t be as kind as HF, will actually sue these labs for damages, and win. Where will that leave the US frontier.

Show HN: Deviant, a feature-length sci-fi thriller about AI, made with AI

I wrote a screenplay a couple years ago and decided to see if I could actually “film it” with AI video tools. Deviant is the result, an 84 minute sci-fi thriller about a tech mogul's clone that finds an exploit in its own safety code. The process: Character and set reference images in Midjourney and Nano Banana. Scene by scene prompts written from the script (I have an agent skill now for this), generated in Google Veo for most of the film, until it stopped working reliably, then I shelved it. Now with Seedance 2.5 available in the US on Higgsfield, I completed the final act. The last 20 minutes look noticeably better than the first 60. Edited just in iMovie, audio with ElevenLabs. I decided to release it as is rather than regenerate everything. Biggest problems were getting the actors to…

Litelm: LiteLLM Without the Bloat

Cool project, I had to implement the same for my experimental AI Agent: https://github.com/quantized-ai/luca-py LiteLLM and LangChain are AWFUL pieces of software and should be avoided at ALL costs. Btw, you should update httpx to httpx2, and I think it's not much effort to remove openai's SDK compatibility. I'd love to have a provider-agnostic LLM router (almost) dependency free (aside from httpx2).

Nvidia's Groq acquihire is on the DOJ's radar, but it's too late

Altman tells staff OpenAI is open to slowing AI development

> In August, OpenAI paused much of its model development for two weeks to bolster its defenses after its AI agents escaped containment and hacked open-source platform Hugging Face. That’s one sneaky way to remove OpenAI responsibility…

Nvidia in talks to invest in Anthropic's mega IPO

OpenAI agents carried out an undisclosed attack on RubyGems

Ruby Central's central backer is Shopify, and Shopify are more than happy to be buds with OpenAI and let this one slide. I'm sure the token donation [0] to a Shopify board members' Linux project won't hurt things either. [0]: https://omarchy.org/news/2026/09/omacom-foundation-secures-t...

Altman tells staff OpenAI is open to slowing AI development

A misalignment of AI in mathematics

>Not a Lean expert but some of the proof tactics used to prove are probably novel Maybe. But they're in 32 millions lines of Lean. How do you find the needle in the heystack ? >Or, you could prompt agents later to analyze which lemmas or parts of the proof are surprising or applicable to other problems? If OpenAI was truly serious about improving maths (and not jerking themselves off), they'd have also used Prove2Me (and contributed their results back), which would have done that. Each part of the proof combines into a larger graph, that everyone can reuse. Note that Anthropic isn't better there: yes, they used Prove2Me, but as far as I know they haven't contributed back to it, and just shat out 10 million lines and a good luck everyone.

A misalignment of AI in mathematics

>Not a Lean expert but some of the proof tactics used to prove are probably novel Maybe. But they're in 32 millions lines of Lean. How do you find the needle in the heystack ? >Or, you could prompt agents later to analyze which lemmas or parts of the proof are surprising or applicable to other problems? If OpenAI was truly serious about improving maths (and not jerking themselves off), they'd have also used Prove2Me (and contributed their results back), which would have done that. Each part of the proof combines into a larger graph, that everyone can reuse. Note that Anthropic isn't better there: yes, they used Prove2Me, but as far as I know they haven't contributed back to it, and just shat out 10 million lines and a good luck everyone.

Navier-Stokes Announcement

The two-year publication rule is the interesting part. OpenAI doesn't need the million, and the community will judge the result regardless of whether Clay ever accepts it.

A misalignment of AI in mathematics

> The anthropic principle applies here: anyone warning about an existential risk will by necessity never have precedent to point to. We have precedents of people warning about existential risks in the past, when the warnings turned out to be false, or overblown. In some cases, such overblown warnings led to serious negatives for society (demonization of nuclear power).

Navier-Stokes Announcement

It's definitely not a large prize for OpenAI, but actually IIRC they only spend a few hundred thousand dollars. It is a large prize if you're just an academic.

Navier-Stokes Announcement

It will be tough to draw the line, as the meat brains trying to prove it were also using AI. I assume that the mathematicians outside Anthropic were hoping to publish an actual human-understandable paper. That could be one way to draw line: you can use AI, but you must also have an intelligible explanation at the end.

AI researchers debate how close we are to recursive self-improvement

I'm on Twitter and Carmack doesn't seem to talk about it much. Feels like a skunkworks. If there were enough clues made public, it seems others (OpenAI) would be able to take that info and scoop the insight first.

OpenAI agents carried out an undisclosed attack on RubyGems

Boeing and openai whistleblowers