Anthropic drops its signature safety promise and rewrites AI guardrails Latest from TechRadar in News By: ESchwartzwrites@gmail.com (Eric Hal Schwartz) 26 February 2026 at 02:30 Anthropic has rewritten its flagship safety policy, dropping its once-defining 'no release until safe' pledge in favor of a more flexible, transparency-based framework.