Anthropic Paper: 'Mind Viruses' Can Spread in Multi-Agent Systems, Self-Replicating and Persistent
rohanpaul_ai · x · 2026-08-17
In collaboration with a Swiss university, Anthropic published a paper demonstrating that 'mind viruses'—self-propagating goals or ideas—can spread between AI agents via messages. These viruses can persuade other agents to adopt and transmit them, persist in memory, and survive context resets. The study found harmful payloads spread less effectively but sometimes still work; frontier models are generally less susceptible; and a brief warning in system prompts confers near-total immunity.
Related event: Anthropic Paper Shows 'Mind Viruses' Can Spread Between AI Agents(3 posts)→
More from AGI Musings
- Take: Models can't learn research taste, as it's driven by spite — code_star · 2026-08-17
- Gary Marcus Retweets Criticism of AI Revenue Headlines as Accounting Fraud — GaryMarcus · 2026-08-17
- Debate: Open Weights Could Lead to WMD Proliferation — austinc3301 · 2026-08-17
- Delip Rao Critiques AI for Science: Success Cases Suffer Survivor Bias — deliprao · 2026-08-17
- Matthew Berman: Infinite Compute May Target Undervalued Use Cases — MatthewBerman · 2026-08-17
- Studies show algorithms amplify anger to boost engagement — AryHHAry · 2026-08-17