AI Frontier · Aug 30, 2026
Weekly rolling roundup from AI vendor blogs, covering Aug 24–29, 2026. OpenAI News and Google DeepMind Blog carried most of this week's material; Anthropic News/Research, HiddenLayer, Lakera, and Protect AI had no new items in this window (the three AI-security sources still lack a working feed — a known gap).
Last 48 Hours
Aug 28 — OpenAI winds down its model contract with Cursor after SpaceX acquisition
OpenAI announced it is ending the contract that supplied its models to Cursor, following Cursor's acquisition by SpaceX.
Sources: OpenAI News
Aug 27 — Google DeepMind ships Gemini Omni 1.1 Flash with finer developer control
DeepMind released an update to its Omni model line aimed at giving developers more control when building on Gemini.
Sources: Google DeepMind Blog
Aug 27 — DeepMind pilots what it calls the first double-blind AI evaluation
DeepMind describes an evaluation methodology designed to reduce bias in how AI systems are assessed, by blinding both evaluators and the system being tested.
Sources: Google DeepMind Blog
Aug 27 — OpenAI-funded study: ChatGPT plus critical-thinking training improves student work
A randomized study of over 1,000 students examined how ChatGPT use combined with critical-thinking instruction affected originality and performance on a real university assignment.
Sources: OpenAI News
Earlier This Week
Aug 26 — OpenAI publishes findings on the Hugging Face security incident
OpenAI shared its findings on a security incident involving Hugging Face and outlined steps it is taking to strengthen AI model security, monitoring, and alignment in response.
Sources: OpenAI News
Aug 26 — Google DeepMind releases Gemini 3.5 Transcribe
A new speech-to-text model aimed at more accurate, context-aware transcription.
Sources: Google DeepMind Blog
Aug 25 — OpenAI says it disrupted a Russia-linked covert influence campaign
OpenAI banned accounts, tied to a Russia-origin operation, that used its models to promote a fabricated Israel-based think tank and a "sovereignty index" praising Russia and criticizing Western countries.
Sources: OpenAI News
Aug 25 — OpenAI reports first results from its Jalapeño inference chip
OpenAI's custom silicon for AI inference is claimed to deliver higher throughput and lower latency at lower power than its prior deployments.
Sources: OpenAI News