White House Withholds Advanced AI Evaluation Framework, Drawing Backlash

According to Axios, the US government has completed an evaluation framework for advanced AI models but decided not to make it public, restricting access exclusively to invited top-tier tech companies. Originating from an executive order on June 2, the framework aims to provide cybersecurity defenders with early access to new models prior to their public release to bolster defenses. This move has drawn heavy criticism from AI safety experts and scholars for its lack of transparency.

已确认

• The White House has completed an evaluation framework for advanced AI models but currently has no plans to release specific details to the public.
• Leading enterprises such as OpenAI, Anthropic, Google, and Meta have been invited to participate in previewing and discussing the framework.
• The framework was developed based on an executive order signed on June 2 and includes an early access mechanism for frontier models.
• Discussion topics involve the definition of covered models, cybersecurity risks, and whether internally deployed models are included.

尚未确认

• It remains unclear whether this voluntary regulatory framework has officially taken effect.

为什么重要

• Refusing to disclose core regulatory details has sparked significant transparency controversies. Commentators such as @dhadfieldmenell and @BlancheMinerva criticized the move as undemocratic "backroom dealing," arguing that such secretive practices will exacerbate risks.
• Security experts like @SOhEigeartaigh pointed out that the secrecy prevents the public from knowing whether the framework covers responses to recent safety incidents involving OpenAI and Hugging Face, thereby raising concerns about the effectiveness of AI regulation.

2026-08-03 ~ 2026-08-05 · 8 related posts

Full story(2 episodes)→

• Episode 1: Trump Administration Nears Voluntary AI Frontier Model Framework(2026-07-28, 7 posts)
• Episode 2: White House Withholds Advanced AI Evaluation Framework, Drawing Backlash(2026-08-03, 8 posts)

• White House invites OpenAI, Anthropic, Google to discuss AI voluntary framework — TorturedPoet30 · 2026-08-03
• White House AI framework questions turn on covered models, cyber risk and internal deployments — Miles_Brundage · 2026-08-04
• White House Won't Release AI Evaluation Framework, Sparking Backlash Over Transparency — BlancheMinerva · 2026-08-05
• White House Completes Voluntary AI Framework But Keeps It Unreleased — 1a3orn · 2026-08-05
• [source] White House Keeps New AI Evaluation Framework Private, Sharing Only With Tech Giants — TorturedPoet30 · 2026-08-05
• White House Won't Release Advanced AI Model Evaluation Framework Publicly — S_OhEigeartaigh · 2026-08-05
• [source] US Government Reportedly Won't Release AI Model Early Access Framework — neil_chilson · 2026-08-05

1 near-duplicate retellings: dhadfieldmenell