Google Cloud has introduced an AI-powered cybersecurity platform that combines Gemini, Wiz, CodeMender and Mandiant expertise to prioritize attack paths and accelerate vulnerability remediation.
Microsoft has announced Project Perception, an agentic security system that coordinates specialized AI agents to identify weaknesses, investigate threats and strengthen defenses at machine speed.
OpenAI says its expanded Daybreak Cyber Partner Program will bring frontier cyber models into approved partners’ security services under defined scopes, monitoring and human oversight.
Anthropic’s analysis of 832 banned accounts finds that AI-assisted attacks are moving deeper into compromised systems and becoming more autonomous.
Cloudflare has introduced Precursor, a client-side verification system designed to identify automated and agentic behavior across an entire web session instead of relying only on isolated checkpoints.
Anthropic’s expanded Project Glasswing points to a practical workflow for turning AI-assisted vulnerability discovery into controlled remediation.
Anthropic’s July case study on Alberta shows how security teams can combine deterministic scanning, AI-assisted review, automated testing and human approval without handing production changes to an autonomous agent.
A practical checklist for using AI vulnerability agents without giving them uncontrolled access to code, credentials or production systems.
Anthropic’s Fable 5 safeguard disclosure provides a practical framework for testing AI-assisted security workflows without exposing production systems or turning a review into an offensive exercise.
A practical workflow for discovering AI endpoints, tuning detection, reducing false positives and safely enforcing security policies.
Anthropic’s limited preview has a credible defensive case, but its public evidence still stops short of proving production-ready autonomous vulnerability remediation.
Anthropic’s published evaluations position Claude Sonnet 5 as a capable general-purpose agent with meaningful safety improvements, but limited evidence of advanced exploit-development ability.
