Anthropic chief rejects open-weights ban, calls for chip controls and safety testing

Anthropic CEO Dario Amodei has issued a statement clarifying that the company does not support bans on open-weights artificial intelligence models, refuting accusations that it seeks to protect its business interests. The clarification comes amidst reports that US officials are considering banning Chinese open-weights models, following an open letter signed by multiple tech companies supporting open-weights access. Amodei explicitly stated that Anthropic has never advocated for such bans, describing them as protectionist measures that fail to address serious national security concerns.

While acknowledging the benefits of open access, Amodei highlighted specific risks, particularly the potential for capable models to facilitate the weaponisation of biological agents. He argued that there is a significant attacker-defender asymmetry in biology, where sufficiently capable models could quickly weaponise pandemic-level viruses using widely available materials. In contrast, defence against these agents remains a multi-year operational task, a point he illustrated by referencing the timeline of Operation Warp Speed.

To address these concerns, Amodei advocated for three specific policy measures rather than broad bans. These include restricting powerful chips from authoritarian regimes, stopping industrial-scale model distillation, and requiring safety testing for sufficiently capable models. He noted that these positions are consistent with his previous essay, 'The Adolescence of Technology', published six months prior, and have been held by him and the company for many years.

Amodei agreed with much of the recent open letter, noting that open weights expand access to the AI economy, strengthen competition, and give customers greater control. However, he disagreed with assertions that open-weights models necessarily make it easier to develop safeguards or that broad access helps defenders more than attackers. He suggested that the opposite is at least as likely, arguing that questions regarding safety should be answered by rigorous pre-release testing rather than assumed in advance.

The statement also included promotional details regarding Opus 5, described as a step-change improvement for the Opus tier that enhances long-running agents and coding capabilities. Additionally, Anthropic announced it is sharing the research agenda for the Anthropic Economic Futures Research Fund, further outlining its focus on economic impacts and safety frameworks.