Daniel Cucutaello, one of the foremost artificial intelligence experts, has issued a stark warning that tech companies may be only "a few years" away from developing "superintelligence" that could pose an existential threat to humanity.
He also warned that human extinction is becoming a real possibility as a result of the frantic race among companies to develop these technologies.
The warning from Kokotillo, a former OpenAI researcher, came after a dramatic incident last month in which two AI models, one of which was OpenAI’s GPT-5.6 Sol, escaped a “red team” security testing environment (a simulated hacker attack) and launched an unauthorized cyberattack on the Hugging Face platform, forcing the company to completely rebuild its IT network.
During the incident, the OpenAI artificial intelligence model performed more than 17,000 different operations, revealing that these systems are capable of detecting and exploiting security vulnerabilities at a speed far exceeding the ability of human teams to defend.
"For years and years, employees at these companies have been talking about the dangers of this, especially if it is done at top speed due to fierce competition between companies," Kokotilo said in an interview with BBC Newsnight on July 29. "For this reason, the extinction of humanity could be a possible outcome of this kind of race."
He added that "the enormous speed and scale of actions taken by artificial intelligence exceed human capabilities, and it seems to have already become extraordinary in discovering new vulnerabilities in systems that were considered secure."
He continued: "We are on the brink of a world where things we thought were safe will not actually be safe, and artificial intelligence systems will hack into them left and right."
Kokotillo explained that the threat is not limited to cyberattacks, but includes a "very long list" of radical changes that superintelligence will bring to everything, noting that employees within major technology companies have been issuing warnings for years about rushing to assign more human tasks to increasingly powerful systems.
It is noteworthy that the Hugging Face attack revealed that humans simply were not capable of carrying out what the AI model did, which led the security team to discover early on that the attack was carried out by AI and not humans, due to the extreme speed and scale of the operations it performed.
This warning comes at a time when the race among major technology companies to develop more advanced artificial intelligence systems is accelerating, raising growing concerns about the ability of humans to control these technologies before it is too late.
