Original: Simon Willison · 16/03/2026
Summary
A member of Anthropic’s alignment-science team discusses the importance of visceral results to convey misalignment risks to policymakers.Key Insights
“The point of the blackmail exercise was to have something to describe to policymakers—results that are visceral enough to land with people.” — A member of Anthropic’s alignment-science team, as told to Gideon Lewis-Kraus
Topics
Full Article
16th March 2026 The point of the blackmail exercise was to have something to describe to policymakers—results that are visceral enough to land with people, and make misalignment risk actually salient in practice for people who had never thought about it before. — A member of Anthropic’s alignment-science team, as told to Gideon Lewis-Kraus Posted 16th March 2026 at 9:38 pm Recent articles My fireside chat about agentic engineering at the Pragmatic Summit - 14th March 2026 Perhaps not Boring Technology after all - 9th March 2026 Can coding agents relicense open source through a “clean room” implementation of code? - 5th March 2026 This is a quotation collected by Simon Willison, posted on 16th March 2026. ai 1913 generative-ai 1696 llms 1662 anthropic 265 claude 261 ai-ethics 279Related Articles
[AINews] Anthropic accuses DeepSeek, Moonshot, and MiniMax of >16 million "industrial-scale distillation attacks"
Swyx · explanation · 65% similar
[AINews] The high-return activity of raising your aspirations for LLMs
Swyx · explanation · 62% similar
Anthropic and the Pentagon
Simon Willison · explanation · 62% similar
Originally published at https://simonwillison.net/2026/Mar/16/blackmail/#atom-everything.