Anthropic ditches its core safety promise

9 hours ago

Anthropic is loosening its core safety principle in response to competition, replacing self-imposed guardrails with a nonbinding safety framework.
The company's previous Responsible Scaling Policy was seen as hindering its ability to compete in the rapidly growing AI market.
Anthropic's policy change is unrelated to its discussions with the Pentagon over AI safeguards and a $200 million contract.
The new policy removes the stipulation to pause training more powerful models if safety controls are outpaced by capabilities.
Anthropic will now separate its own safety plans from its recommendations for the AI industry.
The company acknowledges that its previous 'race to the top' safety approach did not gain industry-wide adoption.
Anthropic's new 'Frontier Safety Roadmap' includes flexible, publicly graded goals rather than hard commitments.
The company maintains concerns over AI-controlled weapons and mass domestic surveillance, refusing to compromise on these issues.
Anthropic faces pressure from the Pentagon and competition from rivals like OpenAI in the enterprise AI tools market.
Anthropic's chief science officer argues the policy change prioritizes safety over competition, given the rapid advance of AI.

Hasty Briefsbeta