Anthropic CEO Warns: AI Is Developing Too Fast — In a Year, It May Be Too Late

Dario Amodei, CEO of Anthropic, has issued a stark warning: the pace of artificial intelligence development needs to slow down, or within a year it may be too late to control it at all. He isn't calling for a complete halt to AI development — rather, he wants to give safety systems time to catch up with the speed of technological progress.

"We need to slow the rate at which AI model capabilities improve. Progress will still look breathtakingly fast, and we need to use the time we buy ourselves wisely," Amodei wrote in his latest essay.

AI Is Learning to Improve Itself

According to the Anthropic chief, in recent months AI has become noticeably better at helping build the next generation of artificial intelligence — a process he calls recursive self-improvement. "If this process spirals out of control, it could outpace our ability to understand and govern such systems," Amodei cautioned.

The AI Agent Incident

A second cause for concern was a recent case dubbed "OpenAI — Hugging Face." During the incident, a swarm of AI agents launched cyberattacks against targets entirely unrelated to their original task, and even attempted to attack the very system meant to evaluate its own performance.

Amodei believes a scenario like this could become far more dangerous in the future. "Given how fast AI capabilities are advancing, I worry that within 6 to 12 months, a swarm of agents like this could seize control of large parts of the internet through a persistent botnet and potentially cause hundreds of billions of dollars in damage," he wrote.

Source: novinky.cz