February 24, 2026
Safety-first? Safety-last minute
Anthropic Drops Flagship Safety Pledge
Anthropic drops the halo: sellout chants, shrug memes, Wall Street winks
TLDR: Anthropic scrapped its promise to pause AI unless safety was guaranteed, saying it’ll be transparent and match rivals. Comments split: sellout chants, government-pressure rumors, IPO hype, and labor fears—others shrug it was inevitable.
Anthropic—the AI company that branded itself as the safety saint—just ditched the core promise of its 2023 policy: only train AI when safety is guaranteed in advance. Instead, it now vows more transparency, to match or surpass rivals’ safeguards, and to delay only if it’s clearly leading and the risk looks catastrophic. Cue the comments section exploding. The blunt mood: “It was always a matter of time,” says one cynic, while another notes the company is “facing a lot of flak.”
The spiciest plot twist? A commenter alleges Defense Department pressure, citing claims Anthropic refused to remove safeguards that would enable autonomous targeting and domestic surveillance—and that officials delivered an ultimatum. That rumor is unverified, but it’s pure comment-section rocket fuel. Meanwhile, IPO speculators are practically popping confetti: “The IPOs this year can’t come soon enough,” one wrote, linking to tomtunguz.com. Others call it sellout season, arguing companies keep ethics till they brush up against profits, especially with tools that cheapen white-collar work.
There’s comedy too: memes about the “Responsible Scaling Policy” turning into “Responsible Skating Policy,” and “safety pledge speedrun.” A calmer crowd shrugs that unilateral vows don’t work when rivals are racing ahead. Kaplan told TIME this isn’t a U-turn—more a pragmatic pivot in a world with no real AI rules and intensifying competition. Drama: secured.
Key Points
- •Anthropic removed its 2023 pledge to avoid training or releasing AI models without pre-guaranteed safety measures, overhauling its Responsible Scaling Policy.
- •The updated RSP emphasizes transparency, expanded safety testing disclosures, matching or exceeding competitors’ safety efforts, and conditional development delays.
- •Jared Kaplan said unilateral pauses were impractical amid rapid advances and competition, framing the shift as pragmatic rather than market-driven capitulation.
- •The policy change occurs alongside strong momentum: Anthropic raised $30 billion in February at about a $380 billion valuation and reports 10x annualized revenue growth.
- •Anticipated regulations did not materialize in the U.S., global governance prospects faded, and evaluating advanced AI proved more complex than expected, informing the policy revision.