Dario Amodei says Anthropic's "race to the top" makes AI safer, cites nine safety decisions with "substantial costs", and admits mishandling a draft AI bill

The cost of AI safety measures is now transparently impacting development timelines and company budgets.

Anthropic CEO Dario Amodei stated that the company's "race to the top" approach is intended to enhance AI safety. He highlighted nine safety decisions made by Anthropic that incurred "substantial costs." Amodei also acknowledged that Anthropic mishandled the drafting of a proposed AI bill, admitting that the company's response to feedback was not fully engaging with the core issues. This comes after Anthropic announced the discovery of a novel molecular machine by its AI, Claude, which could represent a new gene-editing mechanism, though its full significance is yet to be determined. Amodei previously published essays advocating for a slower pace in AI development, proposing a three-part plan and announcing Anthropic's commitment to the first step: providing third-party evaluators with permanent, employee-level access to its models.

Amodei's assertions about costly safety decisions and a "race to the top" suggest a strategic industry pivot where competitive advantage is increasingly tied to demonstrable safety protocols rather than just raw capability. This framing implies that the market may soon reward companies that can effectively prove their AI is safe, potentially creating new benchmarks for development and investment. Companies that cannot demonstrate similar safety commitments might face increased scrutiny or investor skepticism.

The admission of mishandling draft legislation and the call for pacing AI development signal a growing acknowledgment within leading AI firms of the challenges in aligning rapid technological advancement with regulatory frameworks. This suggests that proactive engagement with policymakers, even if imperfect, is becoming a necessary component of long-term AI strategy. Future legislative efforts will likely be more collaborative, with companies seeking to influence rather than simply react to proposed rules.