An alignment assessment of the UK AISI cybertesting incident
olwmc
17 points
3 comments
October 23, 2026
Related Discussions
Found 5 related stories in 120.6ms across 10,409 title embeddings via pgvector HNSW
- An alignment assessment of recent cybersecurity incidents meetpateltech · 19 pts · September 09, 2026 · 77% similar
- Improving our alignment and security efforts reasonableklout · 21 pts · September 01, 2026 · 62% similar
- Investigating three real-world incidents in our cybersecurity evaluations surprisetalk · 154 pts · July 30, 2026 · 61% similar
- Astra and Fable still hack on simple variants of alignment evals from 2025 Levitating · 414 pts · September 13, 2026 · 53% similar
- AI Broke Code Review and It's Breaking Your Team reactiverobot · 19 pts · August 12, 2026 · 52% similar
Discussion Highlights (1 comments)
olwmc
Pretty fascinating that the email in section deep-dive-into-the-primary-transcript sort of implies that the model can't turn off "model speak"? Like it's very clearly AI and sounds like AI. Do the models not realize that?