The Pulse
Palantir CTO Blames Human Error in Iran School Strike
Palantir CTO Shyam Sankar defended military AI and disputed reporting that overreliance on the company’s Maven system contributed to a deadly strike on a school in Minab, Iran. He said a Pentagon report should determine what happened and sa
AI.info Team · Written with AI ·
“Well, we should wait for the final report. But what I can say is that reporting is inaccurate.”
Shyam Sankar, chief technology officer, Palantir
Palantir CTO Shyam Sankar rejected claims that military personnel relied too heavily on AI before a deadly strike hit a school in Minab, Iran. Speaking with CBS News’ Margaret Brennan on “Face the Nation” on October 11, he said the Pentagon’s final investigation should establish what happened, while offering his own view: “the likely cause is human error.”
Sankar disputes the account of Maven’s role
Brennan asked Sankar about Bloomberg’s report that some personnel relied too heavily on AI in Maven, the military platform made by Palantir. Sankar called that reporting inaccurate. He did not describe the evidence behind his conclusion, and said he did not know when the Pentagon would release its final report.
CBS News cited at least 123 schoolchildren killed in the Minab strike. Asked whether the failure involved outdated intelligence or human error feeding information into the system, Sankar again said the final report should come first, then repeated that Bloomberg’s reporting was wrong. He said Palantir had given Congress briefings on Maven’s capabilities, but deferred questions about what went wrong in the strike to the Department of Defense.
Palantir says it added AI checks
Sankar said Palantir built additional AI agents after the strike to double-check more of the work. He described those checks as outside the process Palantir owns, but said the company acted in response to the urgency of the incident. He did not specify what the agents check, how they are tested, or whether the Pentagon has adopted them.
His account places Palantir’s response alongside his argument that AI supports a targeting workflow rather than independently selecting targets. Sankar described the process as identifying targets tied to a military objective, checking them against no-strike lists and reviewing proximity to civilian infrastructure. He argued that software can make those steps more repeatable under time pressure, while acknowledging that human error remains a risk.
A faster targeting process, by Sankar’s account
To illustrate the claimed operational benefit, Sankar said that planning 1,000 strikes in one day would previously have taken two weeks of work by 50 people; with AI, he said, one person could do it in four hours. He presented that as a comparison of planning capacity, not a measure of the Minab operation or an independently verified performance result.
The exchange leaves a sharp disagreement unresolved. Bloomberg’s investigation attributed the strike in part to excessive reliance on Maven, while Sankar rejected that account and predicted that investigators would find human error. His statement that Palantir added AI checks offers one concrete response, but the interview did not establish whether those checks address the failure under investigation. The Pentagon’s final report, which Sankar said he was awaiting, remains the outstanding account of what happened.