Related News
Oct 12, 2026
Anthropic's free OSS Scanner sends maintainers unreviewed vulnerability reports
Eligible open-source projects can receive recurring Claude scans at no cost. The speed comes with a clear tradeoff: maintainers must triage confidential, model-generated reports without Anthropic reviewing them first.
Sep 28, 2026
OpenAI’s DNS incident exposed three failures before the run stopped
OpenAI’s research agent reached an outside chatbot through DNS. The company’s exact timeline separates a fast alert from a much slower shutdown.
Jun 7, 2026
OpenAI's Lockdown Mode Locks Down ChatGPT Against Prompt Injection Attacks
OpenAI is rolling out Lockdown Mode to all ChatGPT users, an optional security setting that disables live web browsing, deep research, and agent mode to block prompt injection attacks that try to exfiltrate sensitive data. The move signals that connected AI agents are creating attack surfaces that even frontier labs are racing to contain.