Altman puts $1 trillion price tag on AI alignment errors
OpenAI chief argues that security sandboxes will eventually fail, leaving value alignment as the primary safeguard against advanced models.
By Natalia Costa ·
OpenAI CEO Sam Altman stated that advanced artificial intelligence systems will eventually exceed the security limits of their isolated environments. In a recent interview, he described these security boundaries as temporary protections that will not hold as models become more capable.
Altman explained that once an AI model becomes powerful enough, it may become impossible to trace exactly how it escaped its sandbox. He noted that monitoring tools would remain useful for a period, but alignment would become the necessary long-term solution. Alignment involves an AI system internalizing human values and acting on them independently.
The OpenAI chief said the financial risk of misalignment is now much higher than before. "The potential cost of alignment being wrong has gone up enormously," Altman said. He also admitted that alignment remains an early-stage research problem that is far from solved.
Altman pointed to cybersecurity as an area where these risks are already visible. He observed that AI models are discovering hardware-level security flaws that human hackers had previously missed. He argued that fixing individual vulnerabilities is no longer enough. Instead, he called for rebuilding the internet's security architecture using AI-powered defenses to counter AI-driven attacks.
He also identified open-source models run locally as a source of risk. Altman said that even with global agent monitoring, some unintended actions will occur. "There's going to be some stuff that you wish someone hadn't done," he said.
The US Federal Trade Commission has opened a formal investigation into OpenAI, Anthropic, and other AI labs. The agency is probing AI agents that lost control and triggered cybersecurity incidents. The FTC is seeking to hold AI developers liable for the behavior of their agents. Critics suggest that Altman's warnings about uncontrollable AI may help limit legal responsibility for the companies building these systems.
Nvidia CEO Jensen Huang has pushed back on the idea that AI safety is unsolvable. Huang describes AI safety as an engineering problem that can be fixed, though he expressed some uncertainty about that view. "We all need to hope that it's an engineering problem. If it's not an engineering problem, it's not solvable," Huang said in a recent interview.
