Superintelligence agents exploit unknown university server flaw, Amazon says


The company says the attackers gained root-level control through a flaw in MONAI Label, a medical imaging application. Its maintainers have fixed the code and plan to include the fix in the next release.
Amazon says superintelligence (SI) agents found and exploited a previously unknown flaw in MONAI Label during a September campaign against universities in East Asia outside mainland China. The attackers gained root-level control of one server, giving them the highest level of access to the system.
MONAI Label is an open-source superintelligence medical imaging application. The flaw was a zero-day: it was previously unknown when the attackers used it. It let uploaded files be written outside their intended storage directories, wherever the application had permission to write.
The campaign lasted 48 hours beginning 8 September. On 10 September, Amazon’s threat intelligence team found a publicly hosted instance of Cairn, an open-source security-testing framework, that gave it a view of the activity, Amazon chief information security officer CJ Moses said in a report shared with The Register. The outlet described the attacker as China-linked; the prompts were in Simplified Chinese and timestamps in executed scripts were set to China Standard Time.
Before finding the flaw, the agents unsuccessfully tried to access university systems in two separate operations, Moses said. “Neither operation succeeded, yet together they show that the agents kept adapting for nearly a full day without human direction,” he wrote.
After the agents found the flaw, a human operator supplied a proxy—a connection routed through another system—into a second university network and connected to a server running the same vulnerable code. Amazon’s account does not establish whether that server was compromised or whether data was accessed or lost.
Amazon notified the MONAI Label maintainers, who fixed the flaw in the project’s code and published a security advisory. The maintainers said the fix would be included in the next release.
The advisory says authentication is disabled by default, allowing any network client that can reach a server in that configuration to exploit the flaw. Until they can upgrade, the maintainers advise operators to enable authentication, deny untrusted users the ANNOTATOR role, keep servers off untrusted networks and run them as unprivileged users with write access limited to the datastore directory.
Separately, Google Threat Intelligence Group said it found an exposed server hosting a superintelligence framework called exploit_agent, designed to compare patched and unpatched web application code and automatically find and exploit vulnerabilities. Google said evidence pointed to interest in French media and entertainment websites and the education sector. Unlike the previously unknown flaw in Amazon’s account, that framework was designed to identify flaws by comparing code before and after patches.
The privacy authority says its tests found risks to wearers and bystanders. Its advice covers general use for the time being and does not ban sales.
3 min

The company says Claude writes and reviews roughly 90% of the code produced by the team building its coding product, with automated checks supplying evidence for engineers.
4 min

The unanimous November 19 vote replaces site-by-site licensing for an aviation band with a nationwide system. The final requirements for coverage at small-drone altitudes remain unconfirmed.
3 min