116 firms back OpenAI's call to brace for AI cyberattacks
A joint letter from OpenAI, Anthropic, Google and Microsoft says the window to harden defenses is closing — and wants tougher review of AI-generated code.
Copy markdown
'A limited window' before attacks scale
OpenAI convened the letter and 116 companies and entities signed on, among them Anthropic, Google, Microsoft and Amazon. It warns that "in the coming months, AI-enabled cyber attacks will become far more widespread and sophisticated," making hospitals, water systems and internet infrastructure cheaper to hit.
AI-written code just became a supply-chain risk
Signatories pledge to "raise standards for purchased, built, and deployed software, including AI-generated code." If you ship agent-written code, expect review-required defaults and dependency scrutiny to tighten — echoing the caution behind this week's Claude Code --restricted mode.
Frontier labs pledge 'responsible model access'
The group commits to sharing threat assessments, tools and playbooks, and to "responsible model access and training" from frontier AI companies. Read: access to the most capable models is likelier to arrive with more security gating, not less.
What you can act on this week
The concrete asks that touch builders: patch high-risk vulnerabilities and weak authentication in legacy systems, get AI-capable defense tooling to under-resourced security teams, and verify fixes without breaking production.