More than 100 technology companies — OpenAI, Anthropic, Google, Microsoft, AWS, Cisco, Cloudflare, CrowdStrike, and Oracle among them — published an open letter today urging a unified defense against AI-enabled cyberattacks, warning that such attacks will become far more widespread and sophisticated in the coming months and that critical infrastructure sits in the blast radius, per TechCrunch and Axios. When the firms building frontier AI issue the warning about frontier AI, the signal-to-marketing ratio deserves attention.
The empirical backdrop arrived earlier in August. CrowdStrike's 2026 Threat Hunting Report documents AI embedded across adversary operations: an 89% surge in AI-enabled adversary activity during 2025, 88% of observed exploitation of vulnerabilities with public proof-of-concept code beginning within 48 hours of release, China-nexus actors moving inside 24 hours, and a North Korean group poisoning trusted AI-framework packages in the software supply chain. Google's threat-intelligence team, meanwhile, detailed Russian espionage clusters stealing accounts by abusing legitimate OAuth sign-in flows — no malware required — and the identity-attack wave kept producing corporate casualties, including the twin breaches at Frontier Airlines attributed in litigation to the merged Scattered Spider collective.
Strip the AI framing and the operational shift is concrete: the interval between a vulnerability's disclosure and its weaponization has collapsed to hours, and the dominant intrusion path is a valid credential, not an exploit. Attackers automated their reconnaissance and their social engineering; defenders whose triage still runs at meeting speed have already lost the timeline. Washington is adjusting to the same reality from the offensive side — an August presidential memorandum authorizes vetted private firms to conduct offensive cyber operations against criminal networks under federal oversight, per TechCrunch and Bloomberg.
Machine-speed offense makes governed machine-speed defense the requirement, and the governed part is not decoration. Autonomous monitoring, correlation, and first-line triage are how a security program keeps pace; auditable data, bounded authority, and humans owning every consequential decision are how it stays trustworthy while doing so. That pairing is the founding design of this platform and its assistant layer.
Governance
AI assistants operate within APEX, using governed, auditable data. They do not bypass authority, workflows, or oversight.
The industry letter asks governments and companies to act inside a limited window. Whatever the policy response, the operational response is available now: instrument the domains you defend, correlate them continuously, and put your humans where judgment matters — because the adversary has already made that division of labor.