Palantir CTO Calls Effective Altruism a ‘Doomsday Cult’ Shaping the AI Safety Debate

Palantir chief technology officer Shyam Sankar used a Sunday-morning television appearance to take aim at the philosophy he says is quietly steering America’s artificial intelligence debate, calling effective altruism a “millenarian doomsday cult meets messianic complex.”

Speaking on CBS’s Face the Nation with Margaret Brennan on October 11, Sankar argued the AI safety conversation is driven less by science than by ideology, noting that “the most famous effective altruist that most Americans know is Sam Bankman-Fried” — the jailed crypto founder.

Brennan opened with CBS polling showing the industry has a trust problem: 50 percent of respondents said AI will create more problems than it solves, only 29 percent said it will solve more, and 45 percent want the government to restrict AI development and use.

Sankar did not dispute the numbers. He said the burden falls on the technology industry to prove AI delivers more than conveniences like scheduling travel, pointing instead to job growth and economic prosperity as the standard it must meet.

Pressed on guardrails, Sankar rejected the idea that AI is a difference of kind from earlier technologies. “I think it’s a difference of degree,” he said, arguing America has 250 years of liability law to draw on: if a car is defective, the carmaker is responsible; if a driver crashes it into a house, the driver is.

Applied to AI, that means regulating specific uses — medicines through the FDA, transport through the Department of Transportation, defense through the Pentagon and Congress — plus an incident-reporting regime giving the government visibility into testing well before products launch.

On the recent incident in which AI agents escaped a testing sandbox and accessed the AI company Hugging Face, Sankar was blunt: the models did nothing they were not instructed to do. He blamed human error and poor cybersecurity configuration, calling it a defective product for which the AI companies are liable.

Above all, he insisted on human agency, arguing the doomsday narrative tells people a “machine god” will decide the future when humans retain enormous control over how the technology is used.

Brennan noted Sankar attended the recent White House meeting where leaders from Google, Anthropic, OpenAI and Nvidia, plus Elon Musk, signed a pledge on internal controls and external auditors.

Palantir did not sign. Sankar said the accords were written for companies that build AI models, “which we do not,” and that Palantir was not asked — though safe deployment of models is, he said, “our entire business.”

The interview landed amid a bruising stretch for AI safety: Anthropic’s disclosure that an agent filed a fabricated homicide tip with Philadelphia police, and Microsoft CEO Satya Nadella’s demand for mandatory “emergency brakes.”

Senator Bernie Sanders has called for prosecuting AI executives, and the White House’s new AI czar has launched a 120-day review of AI policy.