Anthropic Rejects Open-Weights Ban and Pushes Safety Tests

A Narrower Policy Target

Anthropic’s latest open-weights post matters because it separates model openness from the specific risks the company wants policymakers to address. On July 27, 2026, the company’s official X account linked to a full position statement and the post reached about 4.47 million views, with more than 6,500 likes and 850 reposts.

"There’s been a lot of speculation about where we stand on open-weights models."

The linked essay, written by CEO Dario Amodei, says Anthropic has not advocated a categorical ban on open-weights models. It describes non-dangerous open-weight models as a public good while arguing that powerful open releases can be harder to guardrail, monitor, or withdraw after publication. That distinction is important because it moves the discussion away from simple open-versus-closed positioning.

Anthropic names three preferred policy levers. The first is stronger control over powerful chips and chipmaking equipment headed to authoritarian governments. The second is a crackdown on industrial-scale distillation operations that can narrow the gap between restricted compute access and frontier-level model capability. The third is mandatory safety testing for all sufficiently capable models, whether their weights are open or closed.

The account usually posts Claude product updates, safety research, and policy positions, so this tweet is best read as Anthropic trying to define its public stance during a live regulatory fight. The next signal to watch is whether U.S. and allied policy follows this narrower path: compute controls, distillation rules, and pre-release tests rather than a broad prohibition on open weights. Source tweet: Anthropic on X.