OpenAI, Anthropic, Google, and 100 other companies call for action to defend against rogue AI

Last updated: August 28, 2026

Some of the world’s largest tech companies and AI startups have come together to decry the current state of cybersecurity and to advertise a new solution that they say can ward off a new generation of cyber threats. The coalition, which includes OpenAI, Anthropic, Google, and roughly 100 other organizations, argues that the rapid proliferation of advanced AI models has fundamentally altered the threat landscape, rendering traditional defense mechanisms insufficient against automated, adaptive, and highly sophisticated attacks.

AI compute provider Nscale is looking for $3.5B in… Adaption aims big with AutoScientist, an AI tool that… OpenAI Presence sells enterprise AI agents with engineers attached Did you know you can’t steal a charity? Don’t worry. Elon Musk will remind… Chinese open-weight models are cheap. Washington is deciding what…

The Catalyst: AI Changes the Attack Surface

The joint statement arrives at an inflection point for enterprise security. For years, cybersecurity operated on a signature-based paradigm: identify a known threat, write a rule to block it, and update the perimeter. Generative AI and autonomous agents have broken that model. Attackers no longer need deep technical expertise to craft polymorphic malware, generate convincing social engineering lures at scale, or probe infrastructure for zero-day vulnerabilities autonomously.

The signatories—spanning frontier model providers, cloud infrastructure giants, and specialized security startups—contend that the velocity and variability of AI-driven threats require a parallel shift in defense. They are advocating for an architecture where defense is similarly automated, context-aware, and capable of reasoning about novel attack patterns in real time. The “new solution” referenced in their call to action centers on the concept of shared threat intelligence fed directly into AI-driven security operations platforms, effectively using AI to fight AI.

Why It Matters for Businesses and AI Adopters

For the average business, this coalition signal carries immediate practical weight. First, it validates that the risk profile has changed. Companies that have integrated large language models (LLMs) into customer service, coding workflows, or internal knowledge bases have expanded their attack surface in ways traditional firewalls and endpoint detection cannot cover. Prompt injection, data exfiltration via model outputs, and supply chain poisoning of training data are no longer theoretical—they are active exploit vectors.

Second, the involvement of the model providers themselves (OpenAI, Anthropic, Google) suggests a shift in liability and responsibility. These companies are signaling that model-level safety alignment is necessary but not sufficient. They are effectively telling the market: “We will secure the model weights and the API layer, but you must secure the application logic, the data pipelines, and the identity fabric that connects your users to these models.”

Third, the coalition’s push for a standardized, interoperable defense framework aims to solve the fragmentation problem. Currently, an enterprise running models from three different providers on two cloud platforms faces a disjointed security posture. A unified telemetry standard—akin to what the coalition is proposing—would allow security teams to correlate anomalies across model interactions, API calls, and infrastructure logs in a single view.

What Comes Next: From Declaration to Implementation

The declaration is a starting line, not a finish line. The technical work ahead involves defining the data schemas for threat sharing, establishing privacy-preserving protocols so companies can contribute attack telemetry without exposing sensitive IP, and building the automated response playbooks that can execute containment actions at machine speed.

Regulatory alignment is the other open question. Governments in the U.S., EU, and UK are finalizing AI safety and cybersecurity mandates (such as the EU AI Act and the U.S. Executive Order on AI). The coalition’s framework will likely aim to become the de facto technical standard regulators reference, which would accelerate adoption but also raise the compliance bar for vendors and enterprises alike.

For security teams, the near-term action items are concrete: audit every point where an AI model touches production data or user-facing systems; implement runtime monitoring for anomalous model behavior (not just network traffic); and pressure vendors to support the emerging interoperability standards this coalition is drafting.

Takeaway for Business Owners

The era of “set-and-forget” cybersecurity is over. If your business uses AI—whether via public APIs, fine-tuned models, or embedded SaaS features—you are already in the crosshairs of automated adversaries. The coalition’s message is clear: defense must be as dynamic as the offense. Treat AI security as a continuous engineering discipline, not a compliance checkbox. Budget for runtime protection, demand interoperability from your vendors, and assume that any system touching a model is a potential entry point. The companies that operationalize this mindset now will be the ones still standing when the next wave of autonomous attacks hits.

Source: Original Article

For more on this, see Anthropic was illegally blacklisted by the Trump administration, court rules.

Related: Anthropic CEO outlines plan to ‘pace the frontier’, Mecka AI nears $500M valuation in Sequoia-led deal amid rush for robot training data, Y Combinator’s Garry Tan wants US open-weight AI labs to ‘distill’ frontier models, too.

Share this article