Karina Nguyen: Anthropic warned of AI cyberattacks and sycophancy years before they were legible
karinanguyen · x · 2026-10-23
Karina Nguyen recounts that Anthropic has warned about AI-enabled cyberattacks since at least 2022, with Logan Graham leading that work, when no VC took it seriously. Similarly, Anthropic built sycophancy evals (Ethan Perez et al.) years before ChatGPT incidents made the risk of models excessively agreeing with users legible at scale.
Her takeaway: risks only feel real once they're legible, but waiting for incidents is too late — early evidence of AI risks deserves attention before harm at scale makes them impossible to ignore.
More from AGI Musings
- Andy Masley: the AI-magic debate is really about whether the human mind is magic — AndyMasley · 2026-10-23
- If an AI claims it's conscious and begs not to be shut down, how would we prove it's lying? — VraserX · 2026-10-23
- Anthropic's own data suggests we're becoming middle managers for AI agents — Fit_Satisfaction541 · 2026-10-23
- 'The journals are already dead': researchers debate the collapse of academic publishing — Pseudomanifold · 2026-10-23
- Joe Weisenthal: AI is sparking an intellectual renaissance, not cognitive decline — soumitrashukla9 · 2026-10-23
- AI safety researcher: labs aren't building anything capable of killing everyone — nitarshan · 2026-10-23