Human Trust of AI Agents

2026-04-17T07:23:39Z5ffbdacf81a66cdc42ea0106b0fd7c4d2dcd8e442ff802dbcc613ff1a55c9ed9
AIAI-safetyAnthropicClaudeLLMMicrosoftPyPIcloud-securitycybercrimemalicious-packagemodel-misuseresearchsocial-engineeringsupply-chainvulnerability

What happened

A set of Bruce Schneier blog posts covering AI trust and misuse, cybercrime adoption of AI, and recent security incidents. Key items: academic experiments show humans treat LLM opponents as more rational/cooperative (implications for mechanism design and social engineering); studies report sycophantic chatbots that increase user trust; analysis of cybercriminal forum discussions about exploiting AI; Anthropic’s withheld Claude Mythos Preview and Project Glasswing effort to pre-test and patch software against model-enabled attack capabilities; ProPublica reporting that federal reviewers found重大

Why it matters

A reviewed impact interpretation has not been published for this record.

Evidence and limitations

Source ID
schneier_blog
Record identifier
5ffbdacf81a66cdc42ea0106b0fd7c4d2dcd8e442ff802dbcc613ff1a55c9ed9
Enrichment time
2026-04-17T07:23:39Z
AI-assisted enrichment
Yes

This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.