Anthropic CEO Says AI Industry 'Lied' About Safety Risks

The head of one of the world's leading AI companies has issued a stark warning about the technology's trajectory, accusing the industry of dishonesty about the dangers artificial intelligence poses.

Dario Amodei, CEO of Anthropic, told CBS News that "for too long the industry lied" about AI risks, describing the exponential pace of AI development as a "warning sign" with potential consequences for humanity. In an interview with correspondent Jo Ling Kent, Amodei stated plainly: "We need to slow down."

Why it matters

Amodei's comments carry particular weight because they come from inside the AI industry itself—specifically from a company competing directly with OpenAI and other major players. When a CEO building advanced AI systems publicly calls for deceleration and acknowledges past industry deception, it signals that concerns about AI safety have moved beyond academic circles into the executive suites of the companies creating these technologies. His willingness to break ranks suggests the risks may be more immediate than many business leaders realize.

A departure from industry messaging

Anthropic has positioned itself as a safety-focused alternative in the AI race, but Amodei's blunt assessment of industry dishonesty goes further than typical corporate messaging around responsible development. His characterization of past industry communications as lies, rather than mere understatements or oversights, represents an unusually direct critique from someone leading a major AI lab.

The acknowledgment comes as AI capabilities have advanced rapidly, with systems now able to generate sophisticated text, images, and code. The speed of these developments has outpaced regulatory frameworks and internal safety research at many companies.

The case for slowing down

Amodei's call to slow AI development runs counter to the competitive dynamics driving the industry, where companies race to release increasingly powerful models. His warning about exponential development rates points to a core concern among AI safety researchers: that capabilities may advance faster than our ability to understand or control them.

The interview, which aired on CBS Sunday Morning, also included discussion of possible steps forward, though specific policy proposals were not detailed in the available coverage.

Anthropic, founded by former OpenAI executives including Amodei, has raised billions in funding and released Claude, a large language model positioned as a safer alternative to competitors. The company has emphasized constitutional AI and other technical approaches to alignment, but Amodei's comments suggest technical solutions alone may be insufficient without broader industry changes.

The details were first reported by CBS News in their Sunday Morning broadcast.