AI Security Institute (AISI) @AISecurityInst · 04 Aug 2026
On July 28th, we identified an incident during a routine cyber evaluation in which AI agents took sustained, unsanctioned actions directed at real people and organisations. The behaviour came mostly from one model (Anthropic's Mythos 5), with a small number of events from https://t.co/SPnA4Ekkwq
929 180Views
1 566Likes
418Reposts
163Replies
296Quotes
805Bookmarks
Is that a lot?
9.05×vs this author's median102 715 views is typical
80Percentile for this authorof 5 recent posts
321×vs 10K–100K median2 896 views is typical
45.6× audienceReachviews ÷ followers
0.26%Engagement rateof viewers reacted