The BaseEssayLessWrong
AI Security Issues Are Social Engineering Issues
According to the excerpt, the author hiblick argues that capability-based security is part of the fix for AI security problems, says AI-discovered software flaws have reached what the post calls day 42 of the Black October, and expects a safer network afterwards.
Why it matters
The same thing, for different readers.
What it means for you
The author says AI is now finding software flaws at scale and expects many bugs to be fixed, but warns that AI systems bring new weaknesses.
For researchers
Hiblick argues for capability-based security and treats the vulnerabilities of language models as a form of social engineering.
The record
- Questions
- How is AI used in attacks, defence and war? · Are AI companies putting stronger safeguards into practice?
- People
- hiblick (Wrote it) · Peter Eckersley (Mentioned)
- What we read
- We have read only the description the publisher gives. No quotations until someone reads the source itself.
- Found by
- The Radar, run of November 12, 2026, via
lesswrong(tier 1: the source’s own channel or an official record) - Cite this
- LessWrong, “AI Security Issues Are Social Engineering Issues”, November 12, 2026. Entry in the Ctrl AI Base, https://ctrlai.com/base/lesswrong-2026-11-12-ai-security-social-engineering
Our notesLight lane: excerpt only. The post refers to a Black October of AI-discovered flaws; the scale of that claim is the author's and was not checked.