Pwn2Own
At Pwn2Own Ireland 2026 in Cork, researchers exploited OpenAI Codex with a single bug and LiteLLM in two separate attacks. Oracle Autonomous AI Database, Chroma and Dynamo were also exploited. The Zero Day Initiative passes the bug details to the vendors for patching.
Mistral
Mistral AI has opened a public preview of Mistral Large 4, its largest model to date. The mixture-of-experts model, with over a trillion parameters in total and 52 billion active per step, can be tried through the API on Mistral Studio for now. The company says the weights will be released by the en
anthropic
Anthropic has expanded its Cyber Verification Program, which gives vetted security professionals access to its models with fewer restrictions. The program now has three tiers, Defense, Red Team and Specialized, and each includes Claude Mythos 5.1, Opus 5.5 and Sonnet 5.5.
Google DeepMind introduced Gemini 4 Argon, its new frontier model built for long, multi-step work. The model can produce up to 1 million tokens of output in a single response. First access goes to cyber defense teams in the Fairwind Program; no date has been given for wider availability.
anthropic
Anthropic's Frontier Red Team found that Zhipu's open model GLM-5.3 can find browser vulnerabilities and turn them into working exploits, and that its safeguards can be bypassed with simple techniques. The findings match NIST's assessment from mid-September.
Cloudflare
Cloudflare now shows how much of a domain's traffic is protected with post-quantum encryption in HTTP Traffic Analytics, Log Explorer and Logpush. The new fields cover both the visitor connection and the origin server connection.
Cloudflare
Cloudflare cast the latest AI models as attackers and tested its own firewall across 45 scenarios. Human review left 49 findings, which led to new SSRF rules. The company also made Threat Signals, which reads open threat reports with AI, free for every account.
Cloudflare
Cloudflare has applied to become a public certificate authority and signed a deal to buy a GlobalSign root trusted since 2012. Certificates will be free and automated, and its first post-quantum certificates are planned for early 2027.
NVIDIA
NVIDIA has announced the Open Agent Safety Platform to keep AI agents within bounds. Its open-source part, OpenShell, puts every network connection and file access an agent makes under rules; Hugging Face and Anthropic showed their first contributions.
Manifold Security
According to a Manifold Security report, yoursite.com and your-domain.com, often used as sample addresses in documentation, now send some visitors to fake security alerts and investment scams. The two addresses appear in more than 350,000 GitHub files and hundreds of agent skills.
Aikido
Security company Aikido pruned and compressed Z.ai's open model GLM-5.3 to create Altar-1. It is designed to let organisations run security analysis on their own servers without sending source code outside.
Cisco Talos
Cisco Talos released an open-source framework called CAIRN and used it to find malware that takes its instructions from four large language models, with no room for human input.