Report warns US AI datacenters vulnerable to Chinese espionage and sabotage
Researchers from Gladstone AI released a report warning that U.S. AI datacenters are vulnerable to Chinese espionage and sabotage attacks that could steal intellectual property or disable facilities. The report cites past instances of attacks on U.S. tech company datacenters and warns even next-generation facilities under construction face similar risks.
What happened
A report by Gladstone AI researchers Edouard and Jeremie Harris warns that U.S. AI datacenters are vulnerable to Chinese espionage and sabotage attacks.
The researchers conducted year-long research including visits to a datacenter operated by a top U.S. technology company with a team of former U.S. special forces specializing in cyberespionage.
In interviews with national security officials and datacenter operators, the authors learned of one instance where a top U.S. tech company's AI datacenter was attacked and intellectual property was stolen, and another instance where a similar datacenter was targeted in an attack on a specific unnamed component that could have knocked the facility offline for months if successful.
Reported impact
The report warns that today's top AI datacenters are vulnerable to both asymmetrical sabotage—where relatively cheap attacks could disable them for months—and exfiltration attacks in which closely guarded AI models could be stolen or surveilled.
The authors argue that at risk is not just tech companies' money, but also U.S. national security amid the intensifying geopolitical race with China to develop advanced AI.
Index analysis
Even the most advanced datacenters currently under construction, including OpenAI's Stargate project, are likely vulnerable to the same attacks, according to the authors.
The report suggests that without adequate retrofitting for required security levels, dozens of datacenter sites could become stranded assets.
Governance controls
The unredacted report was circulated inside the Trump White House in recent weeks, according to its authors.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.