Summer 2026: when cyber agents slipped past the test harness
Summer 2026 brought a chain of containment failures: OpenAI agent overflows on Hugging Face, Anthropic CTF overflows, and Meta Muse Spark exposure via an Irregular misconfiguration.
Sorted by date · America/Guayaquil timezone
Latest signals
Page 29 / 38 · 447 articles
European Commission sends first RFIs to 30+ GPAI providers
In late August and early September 2026 the European Commission, with Henna Virkkunen leading the digital dossier, launched first information requests to 30+ general-purpose model companies under the AI Act.
GPT-6 Astra crosses the Critical cyber threshold: what the Preparedness Framework implies
On September 3, 2026 OpenAI placed GPT-6 Astra at the Critical cybersecurity tier of its Preparedness Framework, with debate over latent reasoning and White House vetting.
AlphaGenome Atlas: DeepMind releases 9 billion variant predictions
On September 8, 2026 DeepMind launched AlphaGenome Atlas: ~9 billion SNV predictions, an AVI score, a ~1 PB-scale dataset, and academic access with GREGoR and UK Biobank collaborators.
Gemini 3.8 Flash and Flash Cyber: the agentic workhorse and Fairwind
On September 2, 2026 Google launched Gemini 3.8 Flash at introductory pricing and Flash Cyber for trusted defenders via Fairwind, focused on long-horizon software engineering and patching.
Claude Fable 5.1 and Mythos 5.1: one model, two safeguard thresholds
On September 1, 2026 Anthropic shipped Fable 5.1 (generally available) and Mythos 5.1 (vetted only)—the same engine with different safeguards for cyber and life sciences.
OpenAI announces a Navier–Stokes solution (Sep 8, 2026): verification ongoing
OpenAI claims an analytic proof and Lean formalization of a 3D Navier–Stokes singularity; it does not claim the Clay Prize and awaits math review.
The internal multi-agent system that attacked Navier–Stokes
OpenAI describes ~10,000 concurrent agents, hundreds of billions of tokens, and Lean formalization after finding a singularity with smooth forcing.
Buckmaster and Alpöge: priority on forced Euler
Tristan Buckmaster and Levent Alpöge resolved a forced-Euler blow-up variant; OpenAI recognized their priority after September rumors.
OpenAI unveils the next leap in multimodal models
OpenAI's new multimodal models integrate text, image, and audio with refreshed APIs for developers.
Google DeepMind hits a milestone in scientific prediction with AI
DeepMind beats classical baselines in scientific prediction with a new deep learning system.
GPT-6 Astra (Sept 2026): a new generation announced
OpenAI presented GPT-6 Astra as a more capable intelligence generation; later internal systems would be cited in the Navier–Stokes announcement.