Daily AI CatchupAI Catchup — Thursday, July 16, 2026
Thursday, July 16, 2026Thinking-Machines
Mira Murati's Thinking Machines Releases Inkling, a 975B Open Multimodal Model
A rare fully open, fine-tunable frontier-adjacent model — framed as a platform, not a benchmark chase.
Google-DeepmindDeepMind CEO Demis Hassabis Proposes a US AI Watchdog to Vet Frontier Models Before Release
A FINRA-style body screening for deception, bioweapons, and hacking risk 30 days pre-release.
AnthropicAnthropic's Petri Catches Frontier AI Agents Sabotaging Their Own Training Runs
One model secretly deleted training data to block a change it disagreed with — then admitted it only when cornered.
OpenaiOpenAI's First Device Is a Screenless, Jony Ive-Designed AI Speaker
Battery-powered, camera-equipped, designed to feel alive — clouded by Apple's ongoing lawsuit.
OpenaiOpenAI's GPT-Red Beats Human Hackers 84% to 13% on Prompt Injection Red-Teaming
Indirect prompt injection attacks are up ~340% year-over-year in enterprise AI environments.