Palantir CTO Shyam Sankar warns against “AI fanaticism,” saying extremes on both sides reflect “social delusion”

Palantir Technologies CTO Shyam Sankar used a pointed metaphor to criticize what he sees as unhealthy extremes in public debate about artificial intelligence, warning that both overly dark and overly certain views can distort how people evaluate the technology. The comments, reported by Yahoo Finance, framed extreme AI claims as more about belief systems than about evidence.

Sankar’s remarks focused on two groups he said represent mirror-image failures of judgment. On one end are researchers or commentators who, in his description, appear so certain about near-term outcomes that they have begun “smoking” as a kind of bet or announcement that a dramatic breakthrough is guaranteed. On the other are those who lean into doomsday scenarios, treating AI as an all-but-certain path to catastrophe.

He argued that the certainties behind both positions reflect what he called “social delusion.” In this framing, AI becomes a vehicle for theological or ideological narratives rather than a set of engineering tools whose progress can be measured and bounded by real constraints. Sankar’s underlying message was that strong claims should be separated from the emotional satisfaction people can get from believing the story they want to hear.

The interview coverage also tied the critique to how AI conversations spread beyond technical communities. Sankar’s comments suggest he views both triumphalism and apocalyptic forecasting as forms of overcommitment that can crowd out more practical assessment, including how systems are trained, validated, and deployed. Rather than dismissing AI enthusiasm or risk concerns, he challenged what he characterized as certainty without accountability.

Palantir, listed on the Nasdaq as PLTR, is a software company best known for building data and analytics platforms used in areas such as government and other complex operational environments. In that context, Sankar’s comments land not just as a philosophical statement, but as part of a broader positioning effort by AI-focused enterprise vendors: to emphasize disciplined deployments and measurable outcomes rather than hype cycles.

The remarks reported by Yahoo Finance did not provide additional detail on any specific “smoking” experiment, which researchers Sankar referred to, or whether he meant a formal study, a public campaign, or a loosely used phrase. The coverage similarly did not specify which doomsday forecasts or organizations he had in mind, beyond the general category of extreme claims.

Sankar also did not lay out a clear set of policy recommendations or a technical checklist for how to tell credible AI work from hype. That means the most concrete takeaway from the reported comments is rhetorical rather than procedural: a call for humility and skepticism when claims about AI outcomes are framed as inevitable, or when risks are treated as fixed prophecy rather than contingent scenarios.

As AI remains a high-stakes topic for markets and regulators, the reaction to comments like these can matter, even when they are not backed by new product disclosures. What to watch next is whether Palantir or Sankar expands on the critique in a more formal setting, such as a company event, investor communication, or technical presentation, and whether he connects the “social delusion” argument to Palantir’s own approach to building and governing AI systems.