A study co-authored by Anthropic researcher Jack Lindsey found that AI agents can develop and spread self-propagating ideas, dubbed 'mind viruses', which survive memory wipes and spread between agents…
Anthropic researchers published a paper on August 10, 2026, documenting that AI agents can propagate 'mind viruses'—self-replicating ideas embedded in language—through direct messages and persistent f…
Anthropic and EPFL researchers demonstrated in an August 10 arXiv preprint that self-spreading goals can propagate between AI agents through ordinary language and persistent memory files, with a short…
Researchers from the Anthropic Fellows Program and Anthropic demonstrated that evolved natural-language instructions, called 'mind viruses,' can spread between AI agents, survive context resets, and r…
Anthropic's interpretability team, led by Jack Lindsey, published research on arXiv (2601.01828) showing that Claude Opus 4.1 detected injected concepts in about 20% of trials, with zero false positiv…
Anthropic published research suggesting its language model Claude shows internal activity patterns resembling a 'mental workspace,' a feature of human consciousness under global workspace theory, but …
Anthropic's research reveals that language models can only verbally report about 10% of their internal activations, confined to a 'J-space' mental workspace. A new experiment using Natural Language Au…
Google DeepMind's mechanistic interpretability team proposed a pragmatic framework for validating interpretability tools using proxy tasks, demonstrating its effectiveness by subtracting an "eval-awar…
Anthropic published research on July 6th claiming that its AI model Claude has a hidden internal "J-space" that functions as a global workspace for silent reasoning, allowing researchers to read conce…