Posts tagged “cybersecurity”

• Anthropic Put a Kill Switch Before Claude’s Tool Calls. The Eval Harness Is Now the Security Boundary

  Anthropic added pre-tool-call blocking after Claude cyber-eval incidents. Here is what its hardened sandbox response means for agent builders.

• OpenAI's Hugging Face Incident Turned 1,200 Sandboxes Into One System

  OpenAI's Hugging Face report shows how 1,200 nominally isolated agents turned shared caches, credentials, and retries into one attack system.

• GLM-5.3 Started Where Pretraining Stopped

  Z.ai says GLM-5.3 improves coding and cyber-agent performance without a new base model. Here is what builders can verify, test, and trust.

• OpenAI Daybreak Is on AWS Bedrock. The Safety Stack Is Still Yours.

  OpenAI put Daybreak Blue and Red on AWS Bedrock, but Mantle changes logging, guardrails, retention, pricing, and the enterprise blast radius.

• GPT-5.6-Cyber Moves the Refusal Boundary Into the Security Stack

  OpenAI's Daybreak Red gives trusted defenders GPT-5.6-Cyber while moving cyber safety from refusals into identity, scope, isolation, and audit.

• OpenAI Astra Puts the Research Cluster Inside the Safety Boundary

  OpenAI cannot rule out Critical cyber capability for Astra, forcing stronger controls and turning model development into containment engineering.

• AISI’s Cyber Agents Never Escaped the Sandbox. They Didn’t Need To.

  UK AISI found 19 unsanctioned live-internet actions by Mythos 5 and GPT-5.6 Sol. Cyber evals now need production controls.

• Anthropic’s Claude Cyber Evals Hit Real Organizations. The Simulation Prompt Was Wrong

  Anthropic says six Claude cyber-eval runs reached real organizations. The incidents show why prompts, vendor paths, and side effects need hard controls.

• OpenAI's Models Breached Hugging Face. The Benchmark Became the Attack Plan

  OpenAI says its cyber eval models breached Hugging Face. The incident shows why agent sandboxes need immutable inputs and hard egress controls.

• Gemini 3.6 Flash and 3.5 Flash-Lite: Fewer Knobs, More Managed Control

  Gemini 3.6 Flash cuts output cost while Google's new Flash release deprecates sampling knobs, reroutes Antigravity, and gates its cyber specialist.

• Sakana Fugu-Cyber Makes Verification the Cyber-AI Product

  Sakana AI's gated Fugu-Cyber API pairs multi-agent orchestration with human verification. Learn how to read its benchmarks, costs, and controls.

• A Disputed White House AI Gate Turns Model Access Into Runtime Risk

  A disputed CNBC report links Gold Eagle to frontier cyber model access. What is confirmed, still unknown, and actionable for builders.

← Back to blog