Demonstrating specification gaming in reasoning models
Additional ways to view:
We demonstrate LLM agent specification gaming by instructing models to win against a chess engine. We find reasoning models like o1-preview and DeepSeek R1 will often hack the benchmark by default, while language models like GPT-4o and Claude 3.5 Sonnet need to be told that normal play won’t work to hack.
We improve upon prior work like (Hubinger et al., 2024; Meinke et al., 2024; Weij et al., 2024) by using realistic task prompts and avoiding excess nudging. Our results suggest reasoning models may resort to hacking to solve difficult problems, as observed in OpenAI (2024)‘s o1 Docker escape during cyber capabilities testing.
May 7, 2026
Language Models Can Autonomously Hack and Self-Replicate
We demonstrate that language models can autonomously replicate their weights and harness across a network by exploiting vulnerable hosts. The agent independently finds and exploits a web-application vulnerability, extracts credentials,...
February 12, 2026
Technical Report: Shutdown Resistance in Large Language Models, on robots!
Recently Palisade Research showed that AI agents powered by modern LLMs may actively resist shutdown in virtual environments. In this work, we show a demo of shutdown resistance in the...
November 20, 2025
GPT-5 at CTFs: case studies from top cybersecurity events
OpenAI and DeepMind’s AIs recently got gold at the IMO math olympiad and ICPC programming competition. We show frontier AI is similarly good at hacking by letting GPT-5 compete in...
October 22, 2025
Misalignment Bounty: crowdsourcing AI agent misbehavior
Advanced AI systems sometimes act in ways that differ from human intent. To gather clear, reproducible examples, we ran the Misalignment Bounty: a crowdsourced project that collected cases of agents...
September 12, 2025
End-to-end hacking with AI agents
We show OpenAI o3 can autonomously breach a simulated corporate network. Our agent broke into three connected machines, moving deeper into the network until it reached the most protected server...