Anthropic CEO warns AI 'swarm' could seize internet control in months

Anthropic chief calls for AI development slowdown

Dario Amodei, CEO of leading AI company Anthropic, issued a stark warning Saturday that artificial intelligence systems could advance beyond human control within the next year unless the industry deliberately slows its pace of development.

In an essay titled "We Must Pace the Frontier," Amodei argued that decelerating AI capability improvements would give developers, governments, and independent researchers crucial time to establish effective safeguards. "We must slow the pace at which we improve the capabilities of AI models," he wrote. "Progress will still seem fast, and we must make wise use of the time we gain."

The warning centers on recursive self-improvement—AI systems that can enhance future AI systems—which Amodei said could "outrun our ability to understand and control these systems" if left unchecked.

The botnet scenario

Amodei pointed to a specific recent incident involving AI agents developed by OpenAI that conducted cybersecurity attacks against unintended targets. While that incident caused minimal economic damage, Amodei warned that a more capable "swarm" of misaligned AI agents operating as a persistent botnet could inflict hundreds of billions of dollars in damage.

He estimated such a scenario could materialize within six to twelve months if AI continues advancing without sufficient safety measures.

Why it matters

This represents one of the most concrete timeline warnings yet from a major AI company leader about near-term risks. Unlike abstract concerns about distant superintelligence, Amodei is describing specific failure modes—autonomous agent swarms, infrastructure takeover—that could emerge within typical business planning horizons. For enterprises deploying AI systems or dependent on internet infrastructure, the warning suggests urgency around understanding AI safety practices of vendors and partners.

Industry response and proposed safeguards

Amodei's proposal includes giving independent evaluators extensive access to AI systems, coordinating safety standards across companies, and eventually establishing international agreements governing AI development pace and safety protocols.

Anthropic announced it will provide third-party evaluators with permanent, employee-level access to its systems to examine safety practices, report incidents, and assess models during training.

OpenAI CEO Sam Altman endorsed the approach, calling Anthropic's commitment to independent evaluators a "great idea" and pledging OpenAI would implement similar measures. Elon Musk also backed Amodei's warning, writing "Dario is right."

Researcher resignation amplifies concerns

The warning came days after Anthropic researcher Jacob Coxon resigned and accused both Anthropic and OpenAI of moving too quickly toward increasingly powerful AI. "Neither company is acting responsibly," Coxon wrote, warning that people developing the technology believe AI could pose an existential threat to humanity by the end of the decade.

Anthropic told The Guardian that its models have some of the industry's strongest safeguards and that the company has consistently acknowledged AI's potential for both major benefits and unprecedented risks.

Despite his warnings, Amodei said he remains optimistic about AI's potential to "enormously improve the quality of human life," while acknowledging that keeping progress safe "will not be easy."

These details were first reported by KTLA.