Friday Squid Blogging: Squid in Byzantine Monk Cooking

2026-03-07T07:23:43Zc848e526049c5c9c1fba8ea250a9783ceb5cba15ef9faadca353b492e87de9dd
AI-assisted hackingAI-integrity','URL-prompt-parameters'AndroidAnthropicBadeSabaClaudeDoDGambit SecurityGoogle PlayIranIsraelLLM misuseMexican-government-hackOpenAIapp-hackingassassination-assistanceautonomous-weaponsdata-breachmass-surveillancemilitary-AIprompt-injectionpropagandarecommendation-poisoningstate-sponsored-hackingtraffic-camera-compromise

What happened

Collection of March 2026 security items highlighting emergent risks from AI and state cyber operations. Key AI risks: misuse of LLMs (an attacker used Anthropic’s Claude to find vulnerabilities, craft exploits, and automate data theft against the Mexican government), LLM-assisted deanonymization that can re-identify users across platforms, and recommendation/persistence attacks where companies embed hidden prompts in “Summarize with AI” UI buttons to bias assistants (Microsoft found 50+ unique prompts across 31 companies). Policy tensions surfaced as Anthropic refused some DoD uses (mass-surve

Why it matters

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

Evidence and limitations

Source ID
schneier_blog
Record identifier
c848e526049c5c9c1fba8ea250a9783ceb5cba15ef9faadca353b492e87de9dd
Enrichment time
2026-03-07T07:23:43Z
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.