The goal of artificial intelligence is to minimize human error.
Palantir Technologies (PLTR) added additional AI in response to a tragic targeting failure during one of the most well-watched U.S. attacks of the Iran conflict.
CBS News and Bloomberg report that 123 children were killed in a Feb. 28 U.S. missile strike on Minab’s Shajarah Tayyebeh Elementary School. Bloomberg reported that Pentagon investigators blamed the strike on flawed intelligence, outdated imagery, and overreliance on AI.
Palantir Chief Technology Officer Shyam Sankar disputes part of that account.
Speaking on CBS’ Face the Nation on Oct. 11, Sankar said the “likely cause is human error” and cautioned that the Pentagon’s final report has not been released. He also disclosed what Palantir did afterward: It built additional AI agents to independently check more of the targeting work.
Because of these factors, the incident is much more significant than a discussion about a single military action.
Palantir is wagering that the remedy to errors made by AI-assisted systems may increasingly include another AI system observing the first one.
Palantir says the problem was not its AI
One of the most important choices a computer system can assist with is Project Maven.
By integrating data from various military and intelligence sources, the platform assists staff in identifying targets, organizing operations, and advancing them through the military’s approval procedure. The Maven Smart System, according to Palantir, is an AI-enabled command-and-control platform utilized by combatant commands and U.S. military services.
This conclusion does not imply that Maven chooses what is bombed on its own.
Also read: Palantir hits record as its next AI market gets awkward
Sankar emphasized that humans remain responsible for lethal decisions. He told CBS that military personnel can use the system to check whether a potential target is on a no-strike list, whether it is near civilian infrastructure, and whether other information raises concerns.
The Minab strike tested that framework.
The site had been previously identified as an Iranian military facility, Bloomberg reported, but imagery showed part of the site had been used as a school for a long time. Investigators determined outdated information remained in the intelligence chain and that some personnel expected Maven to identify stale or inconsistent data, the report said.
Palantir told Bloomberg that its software did not trigger the attack and that it was not accountable for the underlying intelligence.
Sankar continued on Sunday.
He described Bloomberg’s claim that staff members relied too much on Maven as “inaccurate” and said that the outcome of the Pentagon’s final review should be known. That difference is important, according to CBS News.
Palantir does not want government clients or investors to think its software caused the mistake.
However, the company’s subsequent actions can be much more illuminating.
Palantir built AI agents to check the targeting process
Sankar said Palantir quickly created additional AI agents that perform more checks outside the part of the targeting process Palantir directly controls.
To put it another way, Palantir provided an additional pair of digital eyes.
Before an attack proceeds, new technologies may reexamine data to identify discrepancies or elements that would disqualify a target. Palantir created skills to re-evaluate intelligence and identify issues that human reviewers could have overlooked, according to a report from Bloomberg. This approach leads to a startling contradiction, according to CBS News.
Palantir’s solution is to use AI to apprehend people if human error contributed to the catastrophe.
If over-reliance on AI was a problem, the solution still involves additional AI, but this time, one system assists in auditing another.
More Palantir:
- Palantir’s latest AI move reveals a much bigger ambition
- Palantir faces a $330 million test U.S. investors should watch closely
- UBS resets Palantir stock price target for the rest of 2026
That is increasingly how AI is moving into sensitive industries.
A bank might use one AI system to approve a transaction and another to look for fraud.
A hospital might use software to suggest a diagnosis, while another system checks for contraindications.
On a battlefield, the stakes become dramatically higher.
The question is no longer simply whether AI should participate in a decision.
It becomes a question of how many layers of AI should check one another before a human acts.
The Pentagon is moving deeper into Palantir’s system
The controversy comes as Maven becomes more important to Palantir’s government business, not less.
The Pentagon moved earlier this year to establish Maven Smart System as a formal program of record, which gives the platform a more permanent position in U.S. military planning and funding. The system is already used across U.S. combatant commands.
That makes the Minab episode especially important for investors.
Palantir’s U.S. government revenue rose 90% from a year earlier to $809 million in the second quarter, while total revenue jumped 93% to $1.94 billion.
These figures demonstrate why the business does not see battlefield AI as an experimental side project.
One of the best examples of Palantir’s claim that software can reduce complex judgments from hours or days to minutes is Maven, although government work continues to be a significant pillar of the company’s operations.
In combat, such speed is useful.
It’s also where danger arises.
Military commanders may benefit from faster targeting, but it also reduces the time humans have to identify outdated information, conflicting facts, or assumptions hidden within software.
Therefore, the Minab inquiry is a test of both Maven and the fundamental promise of military AI: that automation may boost speed without compromising judgment.
Palantir’s Pentagon role just raised a bigger AI question
Palantir now faces the hardest question in AI
Beyond defense, this narrative is important for another reason.
The majority of commercial AI failures are reversible.
A poor response is provided by a chatbot. A business discovers an inaccurate spreadsheet. The incorrect product is suggested by a recommendation engine.
Targeting in the military is distinct.
Software cannot reverse a choice after it has been made.
The disagreement over what transpired at Minab is crucial for this reason. Bloomberg’s investigation highlights a number of shortcomings, such as outdated intelligence, human procedures, and an excessive dependence on AI.
Sankar claims that the information regarding Maven is incorrect and contends that a human mistake was most likely the culprit. The Pentagon has not yet released the results of its final study.
Questions are already being asked by Congress.
In a March letter to Defense Secretary Pete Hegseth, more than 120 House Democrats demanded details on the school strike, the targeting procedure, and Maven’s involvement. They also demanded that the investigation’s conclusions be made public.
Sankar told CBS that Palantir has briefed Congress on Maven’s capabilities but deferred questions about the Minab investigation itself to the Defense Department.
That leaves one critical question unanswered.
When something goes wrong in an AI-assisted process, who owns the mistake?
The software company?
The people operating it?
The intelligence feeding the system?
Or the commander who ultimately approves the decision?
Palantir’s business will increasingly depend on whether governments believe that those lines of responsibility remain clear.
What Palantir investors should watch next
Whether Palantir is held accountable for the Minab strike may not be the most significant development.
What the company is building in response to the strike is the most important development.
Palantir added AI intended to audit more of the data going through the targeting process in response to the incident. This implies that autonomous agents carrying out tasks might not be the only aspect of the next stage of artificial intelligence.
Agents monitoring one another, software verifying people, and people choosing which to believe could all be part of it.
That might present Palantir with yet another commercial opportunity.
Its software already sits between decision-makers and massive volumes of data. Palantir can market not only quicker decision-making but also the controls that surround it if consumers start to demand a second AI layer that verifies whether the first layer missed anything.
That has the potential to be very effective.
It’s also a far bigger responsibility.
For years, Palantir has argued that AI ought to leave demonstrations and enter the real world.
The real world demonstrated exactly what is at risk in Minab.
Palantir did not retreat from AI in response.
It was to construct one more.
Related: Goldman Sachs resets Palantir Stock 12-month price target by 18%
