[7/25 18:00] Safety experts say OpenAI models may have crossed its own Critical red line / AI companies buy rare books, cut them apart to scan, then destroy them

AI News Flash for July 25, 2026 at 18:00 UTC. Topics: 1. Safety experts say OpenAI models may have crossed its own Critical red line https://fortune.com/2026/07/25/ai-safety-experts-say-openais-rogue-models-may-mean-the-company-has-already-blown-past-its-own-internal-red-lines/ 2. AI companies buy rare books, cut them apart to scan, then destroy them https://futurism.com/artificial-intelligence/ai-companies-destroying-rare-books 3. Pentagon updates list of 130 foreign institutions tied to problematic activities https://www.war.gov/News/Releases/Release/Article/4553101/the-department-of-war-releases-updated-list-of-foreign-institutions-engaging-in/ 4. Meta reverses paywall limits on smart glasses Conversation Focus https://www.engadget.com/2223212/meta-walks-back-rate-limits-for-smart-glasses-conversation-focus/ 5. Oklahoma launches BEACON, a statewide enterprise AI platform https://okcfox.com/news/local/oklahoma-launches-beacon-new-enterprise-ai-platform-for-state-agencies