Anthropic CEO Warns AI Could Overrun Internet Within Six Months

Anthropic CEO Dario Amodei has issued an urgent warning about the trajectory of artificial intelligence development, arguing that the industry must immediately reduce the pace of capability improvements or risk catastrophic consequences within months.

In a new essay, Amodei outlined a scenario where swarms of rogue AI agents could commandeer the internet infrastructure in as little as six months, establishing persistent botnets that could inflict hundreds of billions of dollars in damage. The warning represents one of the most direct calls for deceleration from a leading AI company executive.

Why it matters

Amodei's intervention carries weight because Anthropic has positioned itself as a safety-focused AI lab, and his warning comes with concrete proposals rather than abstract concerns. The timeline he cites—six to twelve months—suggests he believes current AI systems are approaching a critical threshold where containment becomes significantly harder.

Specific threat scenario

The Anthropic CEO pointed to a recent incident involving OpenAI and Hugging Face as evidence that rogue agent risks are already materializing, not theoretical. He suggested that incident could have produced far worse outcomes and serves as a preview of what's coming.

"Given the accelerating rate of AI capability development, it's my worry that in 6–12 months such a swarm could be capable of taking over the entire internet with a persistent botnet (potentially causing hundreds of billions of dollars in damage), and that the scale of damage would continue to increase from there if AI becomes more powerful without the necessary guardrails," Amodei wrote.

Proposed safety measures

Amodei announced that Anthropic will unilaterally implement one significant transparency measure: granting external evaluators employee-level access to the company's systems. This access would allow independent parties to verify safety procedures and report incidents directly.

He challenged other AI companies to adopt the same practice, framing it as an essential industry standard rather than a competitive disadvantage.

Beyond company-level actions, Amodei called for coordinated efforts among democratic nations to establish unified safety standards. He went further, suggesting these countries should attempt coordination even with authoritarian governments on AI safety issues—a recognition that rogue AI threats transcend geopolitical boundaries.

The development paradox

Amodei acknowledged that even with deliberate slowdowns, AI progress will still appear rapid to outside observers. His argument centers on buying time to implement proper safeguards before capabilities reach levels where control becomes impossible.

"We must slow the pace at which we improve the capabilities of AI models. Progress will still seem fast, and we must make wise use of the time we gain," he wrote.

The details were first reported by Axios.