Update: Insurers Weigh D&O Exposure for Altman and Amodei, FT Reports — EU MEPs Cite His Liability-Framework Call

Updated October 7, 2026: Two October 6 reports pick up the liability thread this piece left open. The Financial Times, as relayed by secondary outlets, says insurers are examining directors-and-officers exposure for Altman and Amodei. Separately, Politico Europe reports EU lawmakers citing Altman’s Decoded remarks to push Brussels to revive a shelved AI liability proposal. The original October 4 daylight story follows below.

October 6–7: liability moves toward pricing and legislation#

Two independent strands landed the same day. Neither is a filing, a priced policy change, or new EU law.

Strand A — insurers and D&O (Financial Times, via relays)#

The Financial Times reported that insurers are bracing for multimillion-dollar claims tied to AI agents acting outside their developers’ intended controls, and that insurers and lawyers are examining whether executives such as OpenAI’s Sam Altman and Anthropic’s Dario Amodei could face claims under directors-and-officers (D&O) policies. We did not inspect the FT original; the account here follows The Decoder, PYMNTS and Beinsure, which all cite the FT — one origin, not three.

Verisk’s Tim Rayner told the FT, per PYMNTS: “Ultimately, the OpenAI CEO is liable [for the Hugging Face incident], because there’s an absence of control in their business.” Broker Aon, per The Decoder’s FT relay, has reviewed more than 300 AI-related legal cases and flagged potential exposure across cyber, intellectual-property and technology-failure lines; there is no case law yet. Hiscox CEO Aki Hussain said it is too early to know how US courts will treat agent liability.

That lands against the Hugging Face incident technical report and the broader rogue-agent disclosure wave OpenAI is still working through (100+ organisations). Italy’s dlgs 160 criminal-liability statute remains a separate national track.

Strand B — EU MEPs and Altman’s Decoded quote (Politico Europe)#

Politico Europe (October 6 evening Brussels time) reports that members of the European Parliament are pressing the Commission to bring back the AI liability proposal it withdrew last year, after Altman told Politico’s Decoded (interview released October 5): “If something goes wrong with our models during training, there’s gonna be some version of that we need to be responsible for… There’s gonna need to be a liability framework for companies.”

Axel Voss, who led the file in Parliament, called it an “overdue moment of honesty.” Brando Benifei called the withdrawal “a serious mistake” and said “frontier models and high-risk systems need strict liability.” That is Benifei’s position — not Altman’s. Altman described a liability framework for training harms; he did not endorse strict liability in the material we read. EU tech chief Henna Virkkunen said last week she would look into potential “loopholes” while stressing the need to be “very careful”; the Commission did not respond to Politico. CEPS research director Artur Bogucki expects any framework Altman backs to come “with many exceptions.”

The two strands answer the same question — who pays when agents cause harm — from opposite institutions. They do not imply insurers reacted to Altman, or that MEPs were answering the FT.

October 4: the original daylight interview#

Sam Altman picked an awkward week to argue for tolerating harm, and he did it anyway. In an interview for Politico’s new Decoded newsletter, conducted by senior tech reporter Brendan Bordelon, the OpenAI chief said: “We believe that the world should accept some bad things happening for the benefits of this technology and people having the agency.”

Asked where he diverges from Anthropic and Dario Amodei, Altman said, “I think there’s a lot of daylight.” Politico’s Forecast newsletter rendered it as “clear daylight.” Read in full, though, the gap is about everyday harms and who controls access, not catastrophe, and it sits on top of a policy record that has been converging.

What he said#

Per the text of Politico’s Decoded article as republished by Jingletree (credited to the Axel Springer Global Reporters Network), Altman said he understands, but disagrees with, the view that the technology is so dangerous “that a single lab in San Francisco should have it and make sure nothing bad happens, and kind of figure out how to dole out the benefits,” calling that “a completely unacceptable tradeoff” at odds with OpenAI’s “lighter touch regulatory stance.” He added: “I wouldn’t take a trade of saying like, ‘We’ll make sure there’s no major hacks, there’s no kind of like misuse of this technology, there’s zero scams’ … Because I think people will do tremendously — orders of magnitude more — good stuff than bad stuff.”

Then the qualifier: he does not accept “the really catastrophic risks,” including “a serious loss of control to AI.” Bordelon, in Politico’s Forecast, called that “a pretty critical distinction” and the point where Altman sees himself “very much in alignment with Anthropic.”

Where the daylight actually is#

The Decoded article itself notes the convergence: Altman agreed with Amodei’s call last month to slow development of the most advanced models; OpenAI has followed Anthropic in endorsing stricter state safety laws; and its lobbyists recently backed a bipartisan House proposal to embed outside safety evaluators at top AI companies. Both companies signed the White House frontier accord days earlier.

So the remaining disagreement, on Altman’s own account, is narrower: how much routine misuse (hacks, scams) to accept in exchange for broad access, and whether concentrating frontier capability is itself the bigger risk. Bordelon read the remarks as preparing the public for more incidents, saying Altman “has already told us this is not the end of the disclosures for autonomous OpenAI hacks.” That lands while OpenAI is still notifying organizations about its own agents.

Limits#

• Nothing has been filed, priced or legislated on either strand.
• The insurer strand rests on the FT via secondary relays; we did not read the FT original.
• Rayner’s exact title varies by relay; we name him as Verisk’s Tim Rayner only.
• Altman’s Decoded liability quote concerns harms during training and a liability framework — not “Altman backs strict liability.” Strict liability is Benifei’s line.
• Politico’s original Decoded article page blocked our earlier fetches; October 4 quotes remain via Jingletree republication and Forecast. The liability quote above is from Politico Europe’s October 6 piece, which we read.
• Forecast says “clear daylight”; the Decoded text quotes “a lot of daylight.” We use the Decoded wording as Altman’s and note both.
• Anthropic had not responded publicly in the October 4 material inspected.

Sources#

• Politico Europe: The EU shelved AI liability rules. Sam Altman has revived the debate (October 6, 2026)
• The Decoder: Insurers brace for millions in claims as AI agents spin out of control (October 6, 2026; FT relay)
• PYMNTS: AI CEOs Could Be Held Liable for Rogue Model Actions (October 6, 2026; FT relay)
• Beinsure: Insurers brace for multimillion-dollar claims over rogue AI agents (October 6, 2026; FT relay)
• Politico Forecast: Sam Altman exclusive: Bad things will happen (Jack Blanchard, October 4, 2026)
• Jingletree (republication of Politico Decoded, credited to Axel Springer Global Reporters Network): Sam Altman said ‘the world should accept some bad things happening’ for the benefits of AI (October 2026)
• Politico: Sam Altman to Decoded: ‘The world should accept some bad things happening’ for the benefits of AI (October 4, 2026; not directly accessible)
• Times of AI: OpenAI Hugging Face incident technical report (August 26, 2026)
• Times of AI: Italy dlgs 160 AI criminal liability (September 30, 2026)
• Times of AI: OpenAI rogue-agent review, 100+ orgs (October 3, 2026)