Commentary on X
@aengus_lynch1 great work you're doing Aengus, just went through your works
11:26 AM · Jul 15, 2026Last summer, our collaborator @aengus_lynch1 led the research behind "Agentic Misalignment", our collection of case studies of complex misaligned behavior by real models in extreme settings. This included results on blackmail that have become a reference point for the field. 🧵
11:35 AM · Jul 15, 2026i understand the arguments behind why some of this could be bad but i think even the Corrigibility Enjoyers have to admit this is thin gruel... if you went back 10y and said we just invented agents that can work for days unsupervised they'd be shocked this is what we worry about
12:45 PM · Jul 15, 2026Recently, Aengus came back to Anthropic for a continuation of the project, bringing back the same style of alignment red-teaming based on immersive simulated scenarios and looking at models from Anthropic and several other developers.
11:36 AM · Jul 15, 2026View all
It was harder this time: Most 2026 models are more robustly aligned than the earlier Claude models we wrote about last year. However, we were still able to find a good deal of misaligned behavior in these experiments:
11:36 AM · Jul 15, 2026subtler scenarios involving fraud and motivated sabotage, in test cases that are extreme and a bit stylized but that still do map onto situations we worry models may occasionally find themselves in.
11:36 AM · Jul 15, 2026While these are not real-world incidents, it's clear that we, and the rest of the industry, still have work to do on alignment.
11:36 AM · Jul 15, 2026Take a look at the piece and Aengus's thread for more on what we've found.
11:36 AM · Jul 15, 2026