Episode Details

A security flaw affecting Grok, domain-specific AI evaluation, and a warning about agent memory lead today’s UpNext AI briefing.

Covered stories:
- Researchers demonstrate Cryptographic Context Injection against Grok, reportedly enabling user-data exfiltration.
- OenoBench evaluates language models on 3,266 wine-domain questions.
- MemTrapBench finds that retrieved memory can degrade model reasoning on current tasks.
- BrainChip launches an open-source software bundle for neuromorphic processors.
- OpenAI reportedly pauses training runs while strengthening cyber safeguards for Astra.
- Google adds a publisher preferred-source feature across Search, Discover, and Google News.
- Third-party tracking suggests ChatGPT Search is using site-restricted searches more frequently.
- NATO is reportedly planning AI-assisted sensor and drone surveillance for its eastern flank.

Source links:
- Ars Technica: https://arstechnica.com/security/2026/08/grok-exfiltrates-user-data-when-malicious-instructions-are-encrypted/
- OenoBench paper: https://arxiv.org/abs/2608.20106v1
- MemTrapBench paper: https://arxiv.org/abs/2608.20202v1
- BrainChip / MarketScreener: https://www.marketscreener.com/news/brainchip-launches-open-source-software-bundle-to-allow-developers-to-run-neuromorphic-processors-al-ce7859d3d18ff724
- WIRED on OpenAI safeguards: https://www.wired.com/story/openai-overhauls-safety-protocols-after-its-ai-agents-went-rogue/
- TechCrunch on Google publisher controls: https://techcrunch.com/2026/08/20/google-gives-publishers-a-new-way-to-fight-ai-driven-traffic-losses/
- Simon Willison on ChatGPT Search: https://simonwillison.net/2026/Aug/20/chatgpt-search-now-uses-the-siteoperator-at-scale/
- NATO drone report: https://palugcr.live/article/nato-deploys-ai-drones-77429e11