Open letter urges Amodei to back mandatory open weights over frontier regulation

An open letter addressed to Anthropic chief executive Dario Amodei surfaced on Hacker News's front page on September 12, 2026, and it takes direct aim at the frontier-safety commitments his company announced the same day. The letter, written by Jacob Gold and published on his personal site, accepts that Amodei is sincere about slowing down — and then argues he is backing the wrong mechanism entirely.

According to the letter, Anthropic's newly published essay, "We Must Pace the Frontier," commits the company to embedding third-party evaluators in its development process and asks governments to impose the same requirement on every other frontier lab. Gold writes that OpenAI's Sam Altman endorsed the approach within hours.

The counterproposal: mandatory open weights

Gold's alternative is a single rule: any AI model a company offers to the public must be released as open weights. Internal and research models would be exempt, and governments could still obtain access to unreleased systems. As concrete examples, the letter says Anthropic would have to publish the weights of its model Fable on release day, and OpenAI would face the same obligation for Astra.

The mechanism Gold points to is financial rather than bureaucratic. Frontier development, he argues, is funded on valuations that assume weights remain proprietary. Voiding that assumption, he writes, would shrink the money available for future training runs and slow progress at every lab at once — in his words, "without a regulator deciding anything."

The regulatory-capture critique

The letter's harder-edged claim is that any regulatory route to slowing frontier models ends in capture. Embedded evaluators, compute thresholds and industry coordination with antitrust waivers all have to be drafted with help from the incumbent labs, Gold argues, because nobody else understands the technical details well enough to write them.

Once such rules become law, he contends, they only accumulate: every incident adds a requirement and no requirement is ever removed. Incumbents can staff the compliance teams and lawyers needed to keep up, while each new obligation raises the wall behind them. Regulation meant to slow the frontier companies down, in his framing, instead protects their position and profits.

Why Gold thinks Amodei is the one to ask

The letter builds its appeal on Amodei's record, as Gold tells it: leaving OpenAI to found a safety-first company when staying was the safer career move; years of calling for regulation of his own industry while being labeled a doomer and a regulatory capturer; and support for export controls that shrink Anthropic's own market. Gold also cites Anthropic's $1.5 billion settlement with authors as evidence of a willingness to pay real costs for principle — a figure attributed to the letter and not independently verified here.

Gold closes by invoking Anthropic's structure as a public benefit corporation, whose charter, he notes, commits it to "the responsible development and maintenance of advanced AI for the long-term benefit of humanity." That legal form, he argues, exists precisely for mission-over-profit decisions of this kind, and Amodei, he writes, remains the only person who could propose an open-weights law and be taken seriously.

Why it matters

The letter crystallizes a central fault line in AI policy: whether frontier development can be restrained through government oversight, or whether such oversight will inevitably be written by, and for, the labs it targets. Anthropic's evaluator commitments — quickly embraced by OpenAI, per the letter — represent the regulatory path. The open-weights mandate represents a structural alternative that attacks the funding assumptions underneath the race itself.

The consequences would be industry-wide. A legal open-weights requirement for public models would erode the proprietary moats behind frontier lab valuations, give researchers, smaller firms and governments access to top-tier systems, and shift competition from model ownership toward everything built around open models. It would also, as the letter itself concedes, demand that Anthropic and its rivals surrender a core commercial asset.

Caveats apply: the description of Anthropic's announcement comes through the letter rather than the essay itself, and there is no indication yet of a response from Amodei or Anthropic. But as a piece of public argument, the letter reframes the debate — not regulation versus acceleration, but regulation versus openness.