More than 100 companies, including OpenAI, Anthropic, Google, Microsoft, and Amazon Web Services, published a joint open letter on Thursday, August 27, 2026, warning that AI-driven cyberattacks are about to get a lot worse. The letter, titled “A Call for Collective Action on Cyber Defense,” was hosted by OpenAI and signed by a list that stretches well beyond the usual AI labs: CrowdStrike, Okta, Fortinet, Broadcom, Capital One, Cloudflare, General Motors, IBM, Mastercard, Oracle, Robinhood, Shopify, and Visa all put their names on it, according to reporting from Engadget, Gizmodo, NBC News, and The Wrap.
The pitch is blunt. Frontier AI models are getting good enough to find and exploit software flaws faster than human defenders can patch them, and the letter argues that the industry has, at most, a handful of months to get ahead of it. That framing, and the roster of signatories attached to it, is why this letter is the story dominating tech coverage today, August 30, 2026.
What the letter actually says
The letter’s central claim is that current security practices will not hold up against AI-enabled attacks. Reporting from Gizmodo and The Wrap quotes the letter’s own language: “In the coming months, AI-enabled cyber attacks will become far more widespread and sophisticated as models around the world become increasingly capable.” The signatories go on to name the sectors they consider most exposed: hospitals, water treatment plants, power systems, and the infrastructure that runs the internet itself.
The letter lays out three principles rather than a single fix. First, it tells organizations to accept that their existing security stack was not built for AI-speed attacks. Second, it calls for arming defenders with the same AI capability that attackers will use, rather than leaving offensive AI tooling to outpace defensive tooling. Third, it asks governments and companies to mobilize a coordinated response instead of each acting alone. None of the three is especially novel on its own, but having OpenAI, Anthropic, Google, and Microsoft co-sign the same three-point plan is new.
Coverage from Politico and Bloomberg cites the letter’s framing directly: “We have a limited window to strengthen cyber defenses.” That single line is doing most of the work in headlines today, and it is the reason security teams are treating this as more than a routine industry statement.
Who signed, and why the list is unusual
What makes this letter different from the usual AI-safety open letter is the mix of signatories. It is not just AI labs warning about AI. Payment networks (Mastercard, Visa), a car maker (General Motors), a retail platform (Shopify), and a brokerage (Robinhood) all signed on, alongside the security vendors you would expect (CrowdStrike, Okta, Fortinet) and the cloud giants (AWS, Google, Microsoft). Reports put the total count at “more than 100” organizations, though the exact figure varies by outlet: some put it at 116, others cite 128 or 130. None of those specific counts has been confirmed as the letter’s own official tally, so treat any precise number you see elsewhere with a little skepticism.
The presence of non-tech companies on the list matters for how this gets read. A cybersecurity warning from OpenAI and Anthropic alone would read as AI labs managing their own reputational risk. A warning that also carries Visa’s and General Motors’ names reads more like a cross-industry risk assessment, and that is likely the point. It gives the letter more weight with the policymakers it is aimed at.
| Category | Named Signatories |
|---|---|
| AI labs | OpenAI, Anthropic, Google |
| Cloud providers | Microsoft, Amazon Web Services, Oracle, IBM |
| Security vendors | CrowdStrike, Okta, Fortinet, Cloudflare |
| Financial services | Capital One, Mastercard, Visa |
| Other industry | General Motors, Shopify, Robinhood, Broadcom |
Why now: the timeline behind the warning
The letter did not appear in a vacuum. Security researchers have spent 2026 documenting a steady rise in AI-assisted intrusion attempts, from phishing kits that write and rewrite their own lures to autonomous scanning tools that probe for unpatched services around the clock. Shattered.io has tracked several of these incidents directly, including a wave of DeFi exploits and a rash of cloud identity misconfigurations that attackers found and chained together faster than defenders could respond. The pattern the letter describes, AI compressing the time between vulnerability discovery and exploitation, is consistent with what security teams have been reporting privately for months.
What is new is the willingness of the model developers themselves to say it publicly, and to attach a rough time horizon to it. “Months,” not “years,” is the word choice across multiple outlets’ summaries of the letter, and that is a meaningfully shorter runway than most enterprise security budgets are built around.
The defensive-AI argument, and its obvious tension
The letter’s second principle, arming defenders with AI, is also its most self-serving, and several outlets have noted the irony. Engadget’s coverage was openly skeptical, pointing out that the same companies calling for collective defense are the ones building and selling the models capable of automating attacks in the first place. The letter proposes expanding trusted-access programs that would give security teams early access to powerful models before public release, which would also, not coincidentally, deepen enterprise dependence on the labs writing the letter.
That tension does not make the underlying warning wrong. Offensive and defensive uses of the same technology have always evolved together, from antivirus signatures chasing malware to intrusion-detection systems chasing network attacks. But it does mean the letter reads partly as a policy document and partly as a market-positioning move, and enterprise security buyers should read it with both in mind.
Market and industry impact
The immediate market reaction has centered on cybersecurity vendors named in the letter. CrowdStrike, Fortinet, Cloudflare, and Okta all got a fresh round of analyst attention this week as the letter reframed AI-driven attacks as an imminent, cross-industry problem rather than a niche concern. For a sector that has spent 2026 competing on “AI-powered detection” marketing claims, a joint statement from OpenAI and Anthropic validating the underlying threat model is a useful talking point heading into enterprise renewal cycles.
On the AI lab side, the letter also functions as a pre-emptive answer to regulators. Governments in the US, UK, and EU have spent much of 2026 pressing frontier labs on dual-use risk, the idea that a model capable of finding software vulnerabilities for defenders is equally capable of finding them for attackers. By publishing a letter that names the risk first and proposes a governance framework around it, OpenAI, Anthropic, and Google get to shape the terms of that conversation rather than react to it.
The dual-use problem, explained simply
Dual-use is the term security researchers use when the same capability helps both attackers and defenders equally well. A model that can read a codebase and spot a buffer overflow is doing the same underlying task whether the person prompting it wants to patch the bug or exploit it. That symmetry is the reason this letter reads differently than a typical vendor security bulletin. OpenAI, Anthropic, and Google are not just warning customers about an external threat, they are describing a property of their own products.
Historically, dual-use risk in software tooling resolved itself slowly. Port scanners, fuzzers, and exploit frameworks all started as attacker tools before defenders adopted them at scale, and the gap between the two adoption curves is where most of the damage happened. The letter’s core bet is that AI-enabled attacks are compressing that gap so fast that defenders cannot afford to wait years to catch up this time. Whether that bet pays off depends less on the letter itself and more on whether the funding and coordination it asks for actually shows up.
How this compares to past industry warnings
Open letters from the AI industry are not new. The 2023 letter calling for a pause on giant AI experiments drew hundreds of signatures but no binding commitments and no pause. Statements on AI extinction risk from 2023 drew similar coverage and similarly little concrete follow-through. What separates this letter is scope: it is not asking for a moratorium or a philosophical commitment, it is asking for funding, coordination, and access programs aimed at a specific, near-term threat category. That is a narrower and more actionable ask, which may be why it drew signatures from companies, like Visa and General Motors, that have no obvious stake in AI-safety philosophy but do have a direct stake in not getting breached.
| Prior AI Industry Letter | Year | Core Ask | Outcome |
|---|---|---|---|
| Pause Giant AI Experiments | 2023 | 6-month training pause | No pause; widely ignored by labs |
| Statement on AI Risk (extinction-level) | 2023 | Treat AI risk as a societal priority | Raised awareness; no binding action |
| Collective Action on Cyber Defense | 2026 | Coordinated cyber defense funding and access | Days old; outcome pending |
What security teams should actually do with this
For working security teams, the letter itself changes little day to day, but it does confirm a shift worth planning around. Attack surface reduction, patch cadence, and identity hygiene were already the baseline recommendations before this letter, and they remain the baseline after it. What the letter adds is a signal that the pace of vulnerability discovery and exploitation is compressing, which means the gap between “patch available” and “patch applied” matters more than it did a year ago.
Practically, that argues for prioritizing automated patch deployment over manual review queues for internet-facing systems, tightening default access for AI coding and agent tools inside the enterprise, and treating any AI vendor’s “trusted access” pitch as a genuine risk-reduction option worth evaluating, not just a sales pitch to wave off. Teams that already run tabletop exercises should add an AI-assisted-attacker scenario to the rotation this quarter rather than next year’s budget cycle.
The regulatory angle
The letter’s call for government coordination lands at an interesting moment for AI policy. The EU AI Act’s higher-risk provisions have been rolling out through 2026, and US federal agencies including CISA have been building out AI-specific guidance for critical infrastructure operators. A joint industry letter asking governments to fund and coordinate defense gives regulators political cover to move faster, since the request is coming from the industry itself rather than being imposed on it. Expect this letter to get cited directly in the next round of congressional hearings and EU Commission statements on AI and critical infrastructure.
Expert and company statements
The letter itself is the primary source for most of the language circulating this week. Its authors write that “In the coming months, AI-enabled cyber attacks will become far more widespread and sophisticated as models around the world become increasingly capable,” a line hosted directly on OpenAI’s collective cyberdefense page. Bloomberg’s coverage of the same letter quotes its central urgency directly: “We have a limited window to strengthen cyber defenses” (Bloomberg). The Hill’s reporting captured the letter’s closing appeal to industry and government leaders to commit their “technology, resources, and expertise” to the effort, and its stated goal of turning “today’s AI advances into lasting improvements in security that benefit everyone” (The Hill).
Where this could go wrong
The obvious risk is that this becomes another 2023-style letter: a lot of signatures, a news cycle, and no funding attached. Coordinated cyber defense at the scale the letter describes needs money, and none of the coverage so far mentions a dollar figure or a funding mechanism. Without one, “collective action” risks staying rhetorical while individual companies keep shipping AI capability faster than shared defenses can absorb it.
There is also a coordination problem baked into the signatory list itself. Getting OpenAI, Anthropic, Google, Microsoft, and AWS to agree on a three-point statement is one thing. Getting them to agree on shared technical standards, data-sharing agreements, or joint incident response protocols, the kind of specifics that would actually move the needle, is a much higher bar, and one that competing commercial interests have made difficult in every prior industry-coordination effort in security.
History offers a useful comparison point outside of AI specifically. The Cybersecurity Information Sharing Act of 2015 tried to solve a similar coordination problem by giving companies legal cover to share threat data with each other and with the government. Adoption was slow and uneven for years, and most security teams still describe cross-company threat sharing as clunky at best. If a legally incentivized program took the better part of a decade to gain real traction, a voluntary open letter will need a lot more than good intentions to move faster.
Predictions: what happens next
- Expect at least one US congressional hearing referencing this letter within the next two months, likely tied to critical-infrastructure cybersecurity legislation already in motion.
- Cybersecurity vendors named as signatories, particularly CrowdStrike, Fortinet, and Cloudflare, will lean on the letter in Q4 2026 marketing and enterprise sales pitches.
- At least one more AI lab not currently listed as a signatory will add its name within weeks, given the reputational pressure of being visibly absent from a 100-plus company list.
- Watch for a follow-up announcement detailing an actual funding commitment or joint program; without one, expect the letter’s news cycle to fade within two to three weeks, similar to prior AI open letters.
- Expect increased scrutiny of “trusted access” programs as critics, echoing Engadget’s early skepticism, question whether they function as genuine risk reduction or as a way to lock enterprises deeper into specific AI vendors.
The bottom line
The letter is real, it is broadly signed, and its core warning, that AI is compressing the time between vulnerability discovery and exploitation, matches what security researchers have been seeing throughout 2026. What it is not, yet, is a funded plan. The next few weeks will show whether “A Call for Collective Action on Cyber Defense” turns into actual coordinated spending and shared infrastructure, or whether it joins the list of AI industry letters that generated headlines without generating change.
Frequently Asked Questions
What is the “Call for Collective Action on Cyber Defense” letter?
It is an open letter published on August 27, 2026, and hosted by OpenAI, warning that AI-enabled cyberattacks will become more widespread and sophisticated in the coming months, and calling for coordinated defensive investment from industry and governments.
Who signed the letter?
More than 100 companies, including OpenAI, Anthropic, Google, Microsoft, AWS, CrowdStrike, Okta, Fortinet, Broadcom, Capital One, Cloudflare, General Motors, IBM, Mastercard, Oracle, Robinhood, Shopify, and Visa.
How many companies signed in total?
Reports describe the total as “more than 100.” Some outlets cite figures like 116, 128, or 130, but no single count has been confirmed as an official, agreed-upon tally.
What does the letter ask governments to do?
It calls for coordination of cyber defense at local, national, and international levels, increased funding for defensive capability, and expanded trusted-access programs that give organizations earlier access to powerful AI models.
Is this the first AI industry open letter about safety?
No. Prior letters, including the 2023 call to pause giant AI experiments and a 2023 statement on AI extinction risk, drew wide signatures but little binding follow-through. This letter is narrower, focused specifically on near-term cyber defense rather than broad AI safety philosophy.
Does the letter include a funding commitment?
Not according to current reporting. The letter calls for increased funding and coordination but does not attach a specific dollar figure or funding mechanism.
What should security teams do in response?
Prioritize faster patch deployment for internet-facing systems, tighten default permissions for AI coding and agent tools, and consider adding AI-assisted-attacker scenarios to existing incident response tabletop exercises.
Why did non-tech companies like Visa and General Motors sign?
Their inclusion signals that the risk the letter describes is being treated as a cross-industry operational concern, not just an AI-industry reputational issue, which gives the letter more weight with policymakers.
