Former Google CEO Eric Schmidt warned that advanced artificial intelligence models can be manipulated by hackers, posing serious risks if misused.
CNBC reported that while speaking at the Sifted Summit in London, Schmidt — who led Google from 2001 to 2011 — said AI systems could be stripped of their built-in safety limits, allowing them to generate harmful or even lethal information.
Schmidt described the problem as one of "proliferation," similar to the spread of nuclear capabilities, noting that AI could fall into the hands of bad actors and be re-engineered for malicious use.
"There's evidence you can take models, closed or open, and hack them to remove their guardrails," he said. "In the course of training, they learn a lot of things. A bad example would be they learn how to kill someone."
He credited major AI developers for implementing strong safeguards but said those barriers are not foolproof. "There's evidence they can be reverse-engineered," he added.
Security researchers have long documented vulnerabilities in AI systems, including "prompt injection" attacks and "jailbreaking."
A prompt injection embeds hidden instructions into inputs or web data to trick the model into performing actions it normally wouldn't, such as revealing confidential information or executing harmful tasks.
Jailbreaking manipulates a model to ignore its restrictions and produce banned or unsafe responses.
A notable example emerged soon after OpenAI's ChatGPT launch in 2023, when users created a "Do Anything Now" persona that forced the chatbot to bypass its ethical guidelines.
The manipulated version was able to provide instructions on illegal acts and generate offensive content, highlighting how quickly security measures can be undermined.
Schmidt said the world still lacks an effective "non-proliferation regime" to regulate the spread of AI technology and prevent abuse. However, he expressed optimism about AI's broader potential, calling it "underhyped" despite widespread public attention.
He cited his collaboration with the late Henry Kissinger, arguing that AI represents a form of "alien intelligence" that will eventually surpass human capabilities.
"The GPT series gave us a glimpse of that," he said, noting ChatGPT's rapid growth to 100 million users within two months of its release. "I think it's underhyped, not overhyped, and I look forward to being proven correct in five or 10 years."
Schmidt also dismissed comparisons between today's AI boom and the dot-com bubble, saying investors appear confident in long-term returns.
"They believe the economic payoff will be enormous," he said. "Why else would they take the risk?"
A growing number of businesses are incorporating AI into daily operations, though adoption remains uneven.
According to the U.S. Census Bureau, just 3.8% of American companies report using AI in production, with usage highest in the information sector at 13.8%.
Jim Mishler ✉
Jim Mishler, a seasoned reporter, anchor and news director, has decades of experience covering crime, politics and environmental issues.
© 2025 Newsmax. All rights reserved.