Additional reporting details Hegseth's 90% cuts to civilian harm mitigation teams and Palantir's post-strike software changes
Pentagon: Maven AI overreliance, outdated data contributed to school strike killing 123
The findings put AI-assisted targeting systems and civilian harm mitigation processes under direct scrutiny after a strike killed 123 children. The combination of outdated data fed to an AI system and the near-elimination of human review staff is identified by Pentagon investigators as the causal chain.
When stories like this move, I write about them in an email each morning.
One free email each morning, around 8 AM US Eastern.
Check your email.
I have sent you a confirmation link. One click and you are in.
Not in your inbox? Look in spam or promotions. If it's there, mark it "Not spam" so future editions land in your inbox.
The full picture
Bloomberg reported that Pentagon investigators determined a U.S. airstrike on an Iranian school that killed 123 children resulted from multiple failures: the Maven Smart System was fed outdated intelligence about a site that had been a military compound but had openly operated as a school since 2018, and Centcom personnel relied too heavily on the AI's output. Hegseth cut civilian harm mitigation staffing at Centcom by roughly 90%, reducing that team from 10 to 1, and no civilian harm mitigation member reviewed the Minab site before the strike. Palantir, whose software runs Maven, said it 'is not responsible for the underlying data nor identifying intelligence deficiencies' and denies its software was at fault. After the strike, Palantir built new Maven capabilities to re-review underlying intelligence and flag inconsistencies. Hegseth allegedly buried the error report for months.
How it developed
Bloomberg reports Pentagon investigators found Maven AI overreliance and outdated data contributed to school strike killing 123 children
Sources
- ChatGPT youth harm lawsuits and APA partnership
- State lawsuits over ChatGPT and minors
- California's SB 813 voluntary AI safety framework
- Florida AG's chatbot lawsuit against OpenAI
BackgroundWhen AI doesn't do what we meant
Get the daily email
When stories like this move, I write about them in an email each morning.
One free email each morning, around 8 AM US Eastern.
Check your email.
I have sent you a confirmation link. One click and you are in.
Not in your inbox? Look in spam or promotions. If it's there, mark it "Not spam" so future editions land in your inbox.