Ex-DeepMind comms staffer: we were banned from discussing extinction risk
j_asminewang · x · 2026-09-10
vmaini, a former Google DeepMind communications & policy team member (2018–2022), reveals the company's internal PR playbook — corroborated the same day by a researcher's resignation from Anthropic:
- At GDM, no one at any level was allowed to discuss human extinction risk externally. Researchers were PR-trained to deflect with "engaging in that alarmism isn't useful; that's Terminator-movie confusion; our AI is safe by design," then pivot to health and climate benefits
- Internally, alignment was unsolved, reward hacking was the default behavior of RL agents, and far too few people worked on the problem
- After months of internal advocacy, the policy was updated to only allow positively valenced, comms-friendly content
Quoted poster @hilbertspaess announced their resignation from Anthropic the same day: after three years of pretraining research across OpenAI and Anthropic, they say both companies are "racing straight to self-improving superintelligence and gambling with our lives."
More from Companies & People
- Report: Anthropic building extensive monitoring system to track anti-AI activists — CtrlAltDwayne · 2026-09-10
- Sakana AI partners with Sumitomo to push domestic AI to 100,000 companies — SakanaAILabs · 2026-09-10
- Automattic board forces CEO Matt Mullenweg on leave; CFO Mark Davies named interim CEO — bytebot · 2026-09-10
- Anthropic's Ethan Perez: departing researcher was worried about pace of progress — EthanJPerez · 2026-09-10
- Jensen Huang: AI Will Grow the $100 Trillion Global GDP Toward $500 Trillion — rohanpaul_ai · 2026-09-10
- Ex-Ant exec says he left at its peak to build frontier AI, downplays doom — _arohan_ · 2026-09-10