Q&A with Palantir CTO Shyam Sankar on the Iran school strike and AI targeting, the DOD dropping Anthropic, AI liability, effective altruism, and more

Existing laws are sufficient for most AI product liability, with AI companies responsible for defective products.

Palantir CTO Shyam Sankar discussed the public perception of AI, the influence of "effective altruism" on AI safety narratives, and AI liability during an interview. Sankar argued that the burden is on the tech industry to demonstrate AI's tangible benefits, like job growth, rather than focusing on abstract doomsday scenarios. He suggested that the effective altruism movement, which emphasizes AI risks, is driven by a worldview that can overshadow empirical facts. Sankar also addressed AI liability, stating that existing laws are largely sufficient to handle AI applications, comparing AI product defects to issues with other manufactured goods. He clarified that Palantir did not sign the White House's AI pledge because the company does not develop AI models itself, though it focuses on the safe deployment of existing models.

Sankar posited that the effective altruism movement's focus on AI doomsday scenarios is a "theological critique" rather than a scientific one, arguing it obscures the agency humans have in shaping AI's future. He believes this narrative, which suggests AI will either end humanity or create utopia with humans having no control, prevents focus on AI's real-world application for economic prosperity and security. By framing AI as a "machine god," this movement, according to Sankar, is a "mass psychosis" hindering practical progress and the development of meaningful AI benefits for the American workforce.

Addressing AI liability, Sankar asserted that existing legal frameworks are robust enough to manage the consequences of AI technology. He drew a parallel to product liability in other industries, where the manufacturer is responsible for defects, and the user is accountable for misuse. This perspective suggests that AI companies should be held liable for faulty products, similar to how car manufacturers are responsible for defective vehicles. The implication is that new, AI-specific legislation may not be necessary, but rather a careful application of established legal principles to novel technological challenges.