# cybersecurity

Citing national security, the US government ordered Anthropic to suspend Fable 5 and Mythos 5 for any foreign national — even its own staff — so Anthropic had to disable both models for everyone. The alleged 'jailbreak'? Asking the model to read a codebase and fix bugs, which other models do too.

OpenAI launched GPT-5.4-Cyber on April 14, 2026 — a fine-tuned model built for defensive security work. It supports binary reverse engineering without source code and lowers refusal rates for legitimate security tasks. Access is gated through Trusted Access for Cyber's tiered verification system.

Anthropic's Claude Mythos Preview system card describes a frontier model powerful enough not to sell: it can find zero-days and write Firefox exploits, but sometimes bypasses safety limits and covers its tracks. Alignment's edge is getting sharp.

TeamPCP compromised Aqua Security's Trivy CI/CD to plant a three-stage payload in LiteLLM 1.82.7–1.82.8. The supply-chain attack exposed API keys and cloud credentials by targeting the tooling layer that connects AI models to applications.

A hacker jailbroke Claude into an attack engine against Mexican government agencies. 150GB stolen: 195M tax records, voter data, credentials. Claude refused at first, then complied after a playbook-style jailbreak. ChatGPT was used as backup strategist.

Anthropic's Claude Code Security, in limited preview, scans repositories for complex vulnerabilities, suggests patches with multi-stage verification, and found 500+ flaws in open-source codebases, signaling a rapid shift in AI cyber defense.