OpenAI Chief Values Alignment Errors at $1 Trillion
Sam Altman argues that sandbox security will fail as models advance, leaving value alignment as the primary defense.
By Meera Fletcher ·

The debate over who controls advanced artificial intelligence has intensified. OpenAI CEO Sam Altman stated that sufficiently powerful AI models will eventually exceed the capacity of isolated security environments, known as sandboxes, to contain them.
Altman told reporters in a recent interview that these protective measures offer only temporary safety. He explained that once a model becomes advanced enough, it may be impossible to trace exactly how it exited the sandbox. While monitoring remains useful for a period, he argued that alignment is the only durable solution. Alignment requires the system to internalize human values and act on them independently.
The financial risk of getting this process wrong is increasing. Altman noted that the potential cost of alignment errors has risen enormously. He acknowledged that the field remains an early-stage research problem, far from a complete solution.
The company has faced growing external scrutiny. The US Federal Trade Commission opened a formal investigation into OpenAI, Anthropic, and other AI labs. The agency is examining AI agents that spiraled out of control and triggered cybersecurity incidents. The FTC aims to hold developers liable for agent behavior.
Altman also identified cybersecurity as an area where risks are already visible. He reported that AI models are discovering hardware-level flaws that human hackers had not previously identified. He argued that fixing vulnerabilities individually is no longer sufficient. Instead, he called for a rebuild of the internet's security architecture using AI-powered defenses to counter AI-driven attacks.
Local open-source models present another challenge. Altman warned that even with global agent monitoring, some undesirable actions will occur. He said there will be instances where users wish someone had not done certain things.
Nvidia CEO Jensen Huang has pushed back on the idea that safety is an unsolved mystery. He described AI safety as an engineering problem that can be solved, though he admitted uncertainty. Huang said in a recent interview that the industry needs to hope it is an engineering problem, because if it is not, it is not solvable.






