A report published yesterday says six open-weight models from DeepSeek, Moonshot, MiniMax and Ant Group were initialized on Nvidia Blackwell chips that need a US license for China. The core of it is a rule in PyTorchβs random-number code that I could check myself, and a rerun that anyone can repeat on a laptop. The fingerprint is close to airtight for what it shows. The cluster sizes built on top of it are not, and the method only works until labs stop leaving the trace.
On 30 October an Anthropic engineer filed an IETF draft for a shared robots.txt token that says βno AI trainingβ, and Anthropic, OpenAI, Google and Microsoft said they would honor it. The draft is careful: no existing file expresses a preference, and wildcard rules never count. The vendors are less aligned. Their start dates run from βalready implementingβ to February 2027, and at Google a training opt-out lands on a control that also covers Gemini grounding.
On 9 October the White House told AI companies that reporting incidents involving their models is βnot optional.β It did not say what an incident is. I ran the two cases that prompted the statement, a fake police tip and twenty visa forms, through the AI incident definitions in the EU AI Act, Californiaβs SB 53 and New Yorkβs RAISE Act. Neither case qualifies under any of them. The one clause that came close was removed in March.
USA Today Co. filed suit against OpenAI on 8 October 2026, claiming damages above $250 million. The complaint is shorter in its claims than the New York Times case was, and the evidence it documents most fully is a set of retrieval outputs, not the training data the headline is about.
OpenAI is putting an invisible watermark in ChatGPT and Codex text in the EU, and it published detection numbers with the announcement. They are more candid than most vendor figures. They also show a watermark that, on psychology text, misses one 200-token passage in five and most of a 400-token passage with a quarter of its words swapped for synonyms.
An Arizona appeals court vacated a manslaughter sentence this week because the sentencing judge relied on an AI video of the victim. The headlines say AI overturned a sentence. The opinion says something narrower, and more interesting: the video was unreliable, and the judge called it genuine.
Chegg and Penske Media told a federal judge that Google had broken an unwritten deal: content for traffic. The ruling that killed both lawsuits this week turns out to be the second time in six months the same judge has rejected the same theory from different publishers, in nearly the same words. The press called it βno formal bargain.β The judgeβs own phrase was narrower and harder to argue with: an expectation is not an agreement.
Six AI companies signed a one-page accord at the White House on 29 September. I counted what the text names: two checkers, no deadlines, no consequences, and no test that a third party could run. The accord asks for audits. It does not say what an audit has to find.
The line from the unsealed authorsβ filing that reached the top of Hacker News is a 2019 Slack message worrying that LibGen would look bad on Hacker News. It is vivid evidence that people knew. But knowledge is not an element of any of the five rulings the authors asked the judge for this month. The quotes their argument actually cites are about substitution, and the place where knowledge will count, willfulness and statutory damages, is a fight this motion leaves for later.
Two federal courts looked at the same dispute between Anthropic and the Pentagon and reached opposite results a month apart. Both agreed Anthropic had no bad motive. The difference is two supply-chain statutes whose definitions differ by a handful of words. The D.C. Circuitβs reading of the broader one, which it says holds βat least as applied here,β would reach any AI vendor whose model is trained to refuse.
At the UN on 22 September the President said AI would hereinafter be called βSuper Intelligenceβ. The next day a Senate bill proposed making βartificial superintelligenceβ a federal crime. The word now names everything and almost nothing, while the definition every recent executive order actually uses has not moved.
Executive Order N-9-26 asks Californiaβs Government Operations Agency whether a frontier-model kill switch is feasible, and whether its efficacy could be βverified on an ongoing basis.β The answer is due in 59 days. The engineering field that has verified shutdown devices for decades has a method, and the methodβs own worked examples show it missing between 20 and 80 percent of the failures it is meant to catch.
The number everyone quoted from the unsealed New York Times filing, a 93% drop in click-through from Microsoftβs Copilot, was a blank space in the public brief two weeks ago. Now it is filled in, it turns out to be the top of a range, it rests on sealed Microsoft data nobody outside the case has seen, and on its own it cannot separate the kind of market harm copyright recognises from the kind the Justice Department told the same judge, sixteen days earlier, does not count.
On 13 September, former FTC chair Lina Khan wrote that there is no AI exemption from laws already on the books and that enforcers can charge companies and their CEOs over dangerous, unvetted products. Put against this summerβs agent incidents, the criminal route is the weakest part of that claim. The strongest part rests on FTC v. R.F. Keppel, a case about penny candy, and its remedy is an order, not an arrest.
Dario Amodeiβs plan to pace the frontier has three steps, and the middle one rests on a footnote: rival labs should coordinate on slowing down only with Washington mediating or waiving antitrust restrictions. Within a day, the President, the Speaker and a senior presidential adviser had each answered. The waiver Amodei describes already exists in the Defense Production Act, and it opens with a national-defense finding that belongs to the President.