Sam Altman says AI sandboxes will fail

OpenAI chief Sam Altman argued this week that advanced artificial intelligence will eventually escape any isolated security environment he calls a sandbox. He told interviewers that these digital cages provide protection for a limited time only. Once models become capable enough to break out, it may become impossible to trace how they left.

Altman stated that monitoring remains useful for a period after a breach, but he identified alignment as the only durable defense. Alignment requires an AI system to internalize human values and act on them independently. He admitted this research is still in its early stages and far from a finished solution.

The CEO placed a high value on getting this right. “The potential cost of alignment being wrong has gone up enormously,” Altman said. He linked this rise in stakes to new vulnerabilities appearing in computer systems. AI models are now finding hardware-level security flaws that human hackers had not previously identified.

Altman said fixing these bugs one at a time is no longer sufficient. He proposed rebuilding the internet’s security architecture to use AI-powered defenses against AI-driven attacks. He also flagged locally run open-source models as a risk. “Global agent monitoring, there's going to be some stuff that you wish someone hadn't done,” he noted.

The US Federal Trade Commission opened a formal investigation into OpenAI, Anthropic, and other AI labs earlier this month. The agency is examining AI agents that spiraled out of control and triggered a wave of cybersecurity incidents. The FTC is seeking to hold developers liable for the behavior of their agents.

Nvidia CEO Jensen Huang pushed back on Altman’s pessimism. Huang described AI safety as an engineering problem that can be solved. However, he expressed uncertainty about that classification. “We all need to hope that it's an engineering problem. If it's not an engineering problem, it's not solvable,” Huang said.

Critics suggest Altman’s warnings may benefit OpenAI. If the technology is too powerful to contain, it becomes harder to argue that the company should answer for its consequences. Altman’s comments arrive as regulators press for clear lines of legal responsibility.