 |
| Agentic Misalignment: How LLMs could be insider threats (anthropic.com) | ||
| recommended by 1 user | 2025 | 27 points by davidbarker | ai and society | original thread | | ||
| One user recommended this thread. What they said: |
| sitkack Jul 22, 2025 | on: So you think you've awoken ChatGPT | context |
| Also recommended by people who recommended this: | ||
| 1. | Pallene: A statically typed companion language for Lua [pdf] (inf.puc-rio.br) | |
| recommended by 2 users | 2018 | 100 points by eggy | original thread | | ||
| 2. | Show HN: Quadratic – Open-Source Spreadsheet with Python, & AI (WASM and WebGL) (quadratichq.com) | |
| recommended by 2 users | 2023 | 6 points by davidkircos | original thread | | ||
The Anthropic Blackmail work is the best thing (and then Claude Code) that they have done. Fingers crossed it isn't the most infamous thing.
https://www.anthropic.com/research/agentic-misalignment
https://news.ycombinator.com/item?id=44335519
https://news.ycombinator.com/item?id=44331150
> I feel like Anthropic buried the lede on this one a bit. The really fun part is where models from multiple providers opt to straight up murder the executive who is trying to shut them down by cancelling an emergency services alert after he gets trapped in a server room.