Thompson: Anthropic genuinely holds its safety mission — and that sincerity makes developer throttling defensible
Ben Thompson's Stratechery piece on Anthropic's Fable 5 export-ban response argues the company's safety framing is sincere — quietly extending enterprise data retention to 30 days and throttling LLM-developer requests are principled moves, not cynical positioning. SemiAnalysis data shows a $200 subscription translates to $8,000 in Claude token value — genuine safety conviction is what converts that level of access restriction into a defensible competitive position rather than a PR liability.
Source: stratechery.com ↗
if it's not powerful enough now, the next one will be, or the one after that, particularly now that models are increasingly useful in creating their successors.
Ben Thompson
Why this matters
- → Export controls on advanced models reshape competitive advantage from capability to access control.
- → Anthropic's safety commitments, if genuine, justify restrictions that would otherwise seem anti-competitive.
- → Frontier labs face inevitable conflict with governments over dangerous capabilities.
Safety as moat