OpenAI says it cannot rule out that its upcoming Astra model may have advanced cybersecurity capabilities. The company's Preparedness Framework defines preparedness as capabilities that could create a new threat vector and require safeguards during development. Internal evaluations indicate significant advancements in agentic coding and cybersecurity. OpenAI announced stricter security controls, including isolated testing environments, restricted network and tool access, enhanced model weight protections, encryption, additional monitoring, and sandboxed execution. It pledged to pause internal testing where these controls are absent and to share guidance with third‑party partners for safe high‑risk evaluations. OpenAI plans to implement chain‑of‑thought monitoring that evaluates model reasoning and can interrupt high‑risk activity before release, though it says the safeguards apply only to internal use. Meanwhile, Anthropic announced it is relaxing certain refusal mechanisms in its Fable model to allow more biology‑related prompts, a change intended to reduce competitive pressure from open‑weight Chinese AI models. OpenAI notes that advanced cyber‑capable models should help defenders find and address vulnerabilities before attackers exploit them.