This is an open letter published on August 27, 2026 under the subtitle "An open letter for a global surge in cyber defense," signed by 128 organizations. It is hosted on OpenAI's website under its Security news category; OpenAI is one signatory among 128, and the document is not attributable to it alone.
Press accounts, including Axios of the same date, describe the signatory group as OpenAI, Anthropic and "more than 100 other companies." The list as published runs to 128 organizations.
The argument
The letter opens with a claim about timing: "We have a limited window to strengthen cyber defenses." Its central forecast, which the signatories assert rather than evidence, is that "[i]n the coming months, AI-enabled cyber attacks will become far more widespread and sophisticated as models around the world become increasingly capable." It names hospitals, water treatment plants and "the infrastructure that powers the internet" as the exposed systems.
Against that it sets an asymmetry claim: "Today's AI advances are already giving defenders new ways to fix weaknesses that have accumulated for years," and if the signatories "act decisively," the "defenders' window" can be used to make digital systems substantially more secure. This is the offence–defence balance question that AI and Cybersecurity and Autonomous cyber-agents track; the letter takes the position that the balance currently favours defenders and is time-limited, and does not engage the contrary positions.
Three principles follow. The first is that "status quo security won't be enough," attributing exposure to "[l]ongstanding bugs, excessive permissions, misconfigurations, insecure and unpatched software, weak authentication, and technical debt in legacy systems" and stating that security teams for critical infrastructure "have been historically under-resourced." The second is to "[e]mpower more defenders with cyber-capable AI," on the ground that sharing tools, knowledge and verified fixes "lets one organization's work help protect many others." The third is to "[m]obilize a collective response," with the reasoning that because "[c]yber capabilities are advancing worldwide," this "can be a net positive: no single company should control the future," and a global response requiring new partnerships is therefore necessary.
The four sets of recommended actions
01 — Every organization. Treat cyber defense with "the urgency and coordination of an incident that takes precedence over everything except critical business operations." Fix the highest-risk weaknesses and verify results without disrupting essential services; raise the bar for what is bought, built and deployed, "including AI-generated code"; build in least privilege, strong access controls and defense in depth; "[u]se capable, lower-cost models for broad coverage, and apply frontier capabilities to the hardest problems"; and where a system cannot be patched without disruption, apply and verify compensating controls.
02 — Cybersecurity companies and technology partners. Test defenses continuously against frontier cyber capabilities; strengthen existing tools with AI; make AI-powered defense deployable for critical-infrastructure operators with hands-on help; work with critical-infrastructure supply-chain manufacturers and system integrators to patch and issue interim guidance; and share threat intelligence and tested playbooks. The section proposes three metrics — "how many organizations are protected, how quickly attacks are contained, and whether fixes work."
03 — Governments. Coordinate at local, national and international levels; strengthen existing channels for actionable threat intelligence; fund cyber defense "starting with essential services that lack the staff or budget to act"; expedite the expansion of trusted access programs, especially for critical-infrastructure supply chains; give hospitals, water utilities and local governments access to capable defensive AI, authorized testing and hands-on support; and "[i]mpose costs on attackers."
04 — Frontier AI companies. Provide "responsible model access, significant funding, training, and hands-on support" for under-resourced critical-infrastructure defenders; build observability and security tools; "ensure agentic identities are traceable and accountable"; invest in authorized testing, private disclosure and verified fixes; and share tools, playbooks and threat assessments with governments, security partners and open-source maintainers. The traceable-agentic-identity ask is the operative commitment the letter directs at its own frontier-lab signatories, and connects to the agent-identity work discussed at Autonomous cyber-agents.
What the letter does not do
It sets no commitments. There are no deadlines, no funding figures, no signatory obligations, no reporting requirement and no mechanism for measuring whether any recommended action was taken. The document is a statement of position and a set of proposals, and is easily read as a pledge that it is not.
Composition of the signatory list
The list spans several industries whose simultaneous presence is the document's main structural feature. Frontier labs and large platforms include OpenAI, Anthropic, Google, Microsoft, AWS, IBM, Oracle, SAP, ServiceNow, Snowflake, Adobe, Cisco, Dell, Red Hat, Elastic, Cloudflare, Akamai, Equinix, Okta, Shopify, Figma, GoDaddy, Deutsche Telekom, Lumen Technologies and Uber. Chipmakers include AMD, Arm, Broadcom, Micron and Cerebras Systems.
The security industry is the largest single bloc: CrowdStrike, Palo Alto Networks, SentinelOne, Fortinet, Check Point, Sophos, Zscaler, Tenable, Darktrace, Proofpoint, Snyk, Socket, Corelight, Cyera, ReliaQuest, Obsidian Security, SpecterOps, Trail of Bits, HackerOne, XBOW, Abnormal AI, Aikido, APIsec, Cato Networks, Gen Digital, 1Password, Cloaked, Enclave, Dropzone.ai, Prophet Security, Cogent Security, Cantina Security, Octane Security, RunSybil, RemoteThreat, Tenzai, Clearly AI, depthfirst, Outtake AI, Veria Labs and the Offensive AI Conference.
Financial institutions and payment networks include Citi, Capital One, Mastercard, Visa, U.S. Bank, BBVA, Fifth Third Bank, Nationwide Building Society, DTCC, The Clearing House, FIS, Fiserv, TransUnion, Robinhood, Block, Citadel and ExodusPoint; insurers and brokers include Zurich Insurance Company, Marsh and Oliver Wyman. Professional-services and systems-integration firms include PwC, KPMG LLP, Accenture, Cognizant, Capgemini, EPAM, HCLTech, NTT DATA, Unisys and WWT. General Motors and Flexport sign from outside technology. The only non-profit signatory is the Center for Internet Security.
Several AI companies with existing pages sign without the fact being recorded elsewhere: Hugging Face, Perplexity, Scale AI, Vercel, Harvey, Replit, Cognition AI, Mercor, Glean, Lovable, Factory, Arena, Fleet AI, Calif, Corridor, Cotool, Glow, EXA.ai, Incident.io, TrendAI and Third Moment Research.
Absent from the list as published are Meta, xAI, Nvidia, Apple, Amazon as distinct from AWS, and every Chinese laboratory.
Relationships
- supports: AI and Cybersecurity, Autonomous cyber-agents.
- depends-on: OpenAI, Anthropic.
- related: Financial Services — AI Deployment, Investigating Three Real-World Incidents in Our Cybersecurity Evaluations (Anthropic Frontier Red Team, July 2026), OpenAI and Hugging Face Partner to Address Security Incident During Model Evaluation (OpenAI, July 2026).