Anthropic issues warning: "AI agents could control the Internet within six months"

Dario Amodei, CEO of the AI company Anthropic, has raised an alarm about AI development, stating in an interview with CBS that "AI agents could control the internet within 6 months." The statement comes after the publication of a lengthy post in which Amodei describes the serious dangers associated with the technology.

Amodei's warning expands the debate on AI risks. Key industry figures, including Sam Altman of ChatGPT and Elon Musk, have expressed agreement on the use of external and impartial observers capable of overseeing technology development within companies. Anthropic has also reported blocking attempts to use AI for potential biological weapons research.

The debate on regulating artificial intelligence has been ongoing for years, but it has gained urgency with the rapid development of increasingly powerful models. The turning point was the public launch of systems like ChatGPT at the end of 2022, which demonstrated the revolutionary (and potentially dangerous) capabilities of generative AI. Since then, the technological race among companies like OpenAI, Anthropic, Google, and Meta has intensified.

Anthropic, founded by former OpenAI researchers concerned about safety, has positioned itself as a company focused on "aligned" and safe AI development. However, according to Amodei, the industry has "lied for too long" about the real risks. The specific alarm about autonomous "AI agents" that could control the Internet represents an evolution of the concern: it is no longer just about misinformation or deepfakes, but about systems that could operate online without human supervision, making autonomous decisions with unpredictable consequences. Internal tension over safety has already led to the resignation of researchers who had issued warnings.