Invest6 publishersWidely confirmed3 min readPublished Updated
Huang and Zuckerberg pushed back on Amodei's AI risk warnings at a closed White House meeting
Nvidia's Jensen Huang and Meta's Mark Zuckerberg challenged Anthropic's AI risk warnings at a Sept. 29 White House meeting, the Wall Street Journal reported. Anthropic, chasing a $2 trillion listing, still signed the self-regulation accord that came out of the meeting.
The Investor · Invest desk
Drafted by a language model from the sources cited here and checked against its claim ledger before publication. How we use AISend a correction
What happened
- Amodei did not back down, telling the other chiefs they should be honest with the public about what their models can do and should not understate the risks.
- At the luncheon, Zuckerberg argued against new regulation, saying the safety worries could be handled if the industry adhered strictly to its own principles.
- Anthropic, OpenAI and Google had earlier backed a self-regulatory body modeled on FINRA, but it collapsed after Huang, Zuckerberg and Elon Musk told President Trump they opposed it.
- Anthropic was placed on a blacklist during a dispute with the Defense Department over the scope of military uses of AI.
Compiled by The InvestorSomething wrong?How this is made
Why it matters
- constraint The industry's remaining plans for model evaluations and internal controls need buy-in from chiefs who split in private, and the Journal says that friction could block them.
- decision Anthropic has to weigh the public risk warnings Huang challenged against an IPO that needs better ties with an administration hostile to AI regulation.
- exposure Buyers of an Anthropic listing would take on a company with a Defense Department blacklist on its record, competing with rivals who have the President's ear on AI policy.
Every chief at the table ended the day inside the White House superintelligence accord, which calls for industry-led self-regulation instead of new rules [6]. Outside the room, they kept their language measured [6]. So the fight the Journal describes took place among executives who already agreed there should be no new legislation [1]. It was about how much the industry's own regime should contain. On the most formal version on offer, a body modeled on FINRA, the count was three companies for and three chiefs against [16]. The three against won [5].
The side that won also helped write the text [9]. According to the Journal's sources, Zuckerberg played a decisive role in drafting the accord and was in frequent contact with Trump and Commerce Secretary Howard Lutnick [9]. Meta's relations with Trump hit bottom after it suspended his Facebook account following the Jan. 6, 2021, occupation of the Capitol. Since his second term began, Meta has settled the lawsuit over the suspension and invested in projects the President favors [13].
Anthropic is at an earlier stage of the same repair. Amodei had a one-on-one dinner with Trump at the White House on Sept. 27 [11], two days before the meeting [17]. After the event he told reporters that if the industry and the government work together, "we can win safely" [8], and Trump praised the remark as excellent [14]. The company argued with its peers behind closed doors and kept to measured language in public [6]. It is pursuing a listing and needs to mend its standing with an administration hostile to AI regulation [12][15].
If the Journal is right that the friction could block the industry's plans for self-regulation covering model evaluations and internal controls [7], the accord is left as a commitment against new rules with no agreed rules of its own. If the Sept. 27 dinner becomes a lasting thaw, Amodei's warnings come from a company the White House listens to, and the private argument tilts his way. The whole account also rests on unnamed sources in one newspaper's report, relayed by Seoul Economic Daily [1], so who said what is single-sourced.
I think the private split is wider than the signed accord suggests, with one correction about who said what. Huang's reported move was a question about why Amodei's public warnings are so extreme [2]. The argument against new regulation at the luncheon came from Zuckerberg [4]. The stronger evidence came earlier. Huang, Zuckerberg and Musk opposed even a self-regulatory body run by the industry [5], so the self-regulation they back stops short of anything modeled on FINRA. The view would be wrong if Nvidia and Meta joined a version of the evaluations and internal-controls plan [7] that came with an industry body attached.
What to watch
- Whether the industry's self-regulation plans for model evaluations and internal controls advance or stall over the split with Anthropic.
- Whether the Defense Department lifts Anthropic's blacklist before the company lists.
- Whether Nvidia or Meta publicly disputes the Journal's account of the Sept. 29 meeting.