A coordinated AI slowdown would need an antitrust exemption, Amodei says

Amodei's exemption request is the one item here with a counterparty in government. He said that coordinating with other AI labs would in some cases require an antitrust exemption from the government ^[7]. A company asks for one when the agreement it wants is with competitors and covers something a regulator would otherwise treat as a restraint on output. What a pause restrains is the rate at which each lab ships models.

Altman named four rivals: Anthropic, SpaceX, Google DeepMind and Meta, with others possible ^[1]. That puts five signatures on the smallest version of the deal, OpenAI included ^[17]. Fortune notes that Altman does not get along with either Amodei or Elon Musk ^[16]. He said the discussions were already underway and that a coordinated slowdown might be announced soon ^[2], and he endorsed most of what Amodei had written ^[10].

One deal term did get spelled out. Altman said he would have no problem telling investors that OpenAI had taken actions to prioritise safety that cost them financially, and that OpenAI's investors were warned of that possibility going in ^[3]. The warning came before the money did. He also said OpenAI would not go public this year, giving two reasons: concern about the safety of the latest models, and, more quietly, that the business is not yet in the right place ^[4].

Timing is what makes a slowdown expensive or cheap. Fortune reports Anthropic on the verge of an IPO and OpenAI edging closer to one, and credits that timing with part of the impact of former safety researcher Jacob Coxon's resignation ^[15][14]. A company approaching a prospectus that commits to shipping more slowly is telling public-market buyers that its growth rate is a variable management adjusts. Anthropic's version of the commitment has a named third party in it: independent evaluators permanently on-site to review its safety work, with the nonprofit METR as Amodei's preferred partner ^[8].

The compliance half of the week is harder to size. A watchdog group says OpenAI is violating California's new AI safety law, and an E&Y survey finds half of companies are not following their own AI governance policies ^[11][12]. Both appear as single-line items in the edition's contents list, and Fortune does not name the watchdog group or the provision at issue ^[19].

I would expect the pause to arrive as a joint statement with no dates in it, because that version needs no exemption and no signature from Musk. Two other readings are live. If a lab actually files for an exemption and an agency answers, a real restraint is being negotiated in public, and Amodei's further ask, that democracies seek an international governance agreement with China and other authoritarian states, sets the outer bound of that road ^[9]. If instead the California statute is where the pressure lands, the watchdog's allegation decides more than anything said on a vodcast ^[5]. What would change the first view: a slowdown announced with per-lab scope, dates and evaluators free to publish, or OpenAI filing to list this year after saying it would not ^[4].