US child exploitation investigators are beginning to test artificial intelligence against a problem that artificial intelligence has helped intensify: the rapid spread of synthetic child sexual abuse material, or CSAM.
According to a new government filing, the Department of Homeland Security’s Cyber Crimes Center has awarded a $150,000 contract to San Francisco–based Hive AI. The company’s software can identify whether content was AI-generated, and the trial is aimed at helping investigators separate AI-made material from images that may depict real victims.
Why synthetic CSAM creates a new investigative problem
Generative AI has made it easier to produce child sexual abuse images at scale. The source article says the National Center for Missing and Exploited Children reported a 1,325% increase in incidents involving generative AI in 2024.
For investigators, the central issue is not only the volume of material. Their first priority is to identify and stop abuse that may still be happening. When large amounts of AI-generated CSAM enter the same stream of evidence and reports as material involving real children, investigators need a way to decide which cases require the most urgent attention.
The filing frames automation as a practical response to scale. It says, “The sheer volume of digital content circulating online necessitates the use of automated tools to process and analyze data efficiently.”
That does not make the work simple. A detector must support a high-stakes decision process: whether a piece of content is likely synthetic, or whether it may show a real victim who needs immediate protection. The value of such a system depends on whether it can help investigators prioritize without pulling attention away from children in danger.
What the Cyber Crimes Center is testing
The Cyber Crimes Center investigates child exploitation across international borders. Its contract with Hive AI is heavily redacted, but Hive cofounder and CEO Kevin Guo confirmed to MIT Technology Review that the work involves the company’s AI detection algorithms for CSAM.
Guo said he could not discuss the details of the contract. The filing was posted on September 19, and the trial will last three months.
The key purpose is triage. The government filing says identifying AI-generated images “ensures that investigative resources are focused on cases involving real victims, maximizing the program’s impact and safeguarding vulnerable individuals.”
In plain terms, the government is not describing AI detection as a replacement for investigators. It is presenting the tool as a way to help sort through digital material more efficiently, so investigators can focus on cases where a real child may be at risk.
How Hive AI fits into the existing CSAM defense stack
Hive AI sells tools for creating videos and images, and it also offers content moderation products. Those tools can flag violence, spam, and sexual material, and can even identify celebrities.
For CSAM detection, Hive offers a tool created with Thorn, a child safety nonprofit. Companies can integrate that tool into their platforms. It uses a hashing system, which assigns unique IDs to content already known by investigators to be CSAM and blocks that material from being uploaded.
Hashing tools have become a standard defense for technology companies, but they solve a different problem. They can identify known CSAM. They do not determine whether an image was generated by AI.
Hive’s separate AI-generated image detector is the tool the Cyber Crimes Center will use to evaluate CSAM, according to Guo. He said the detector was not trained specifically for CSAM, but argued that it does not need to be because the signal can generalize across images.
“There’s some underlying combination of pixels in this image that we can identify” as AI-generated, he says. “It can be generalizable.”
Guo also said Hive benchmarks its detection tools for each customer’s specific use case. The National Center for Missing and Exploited Children did not respond to requests for comment on the effectiveness of such detection models in time for publication.
Why Hive received the contract
The government filing justifies giving the contract to Hive without a competitive bidding process. Some of that explanation is redacted, but the visible parts mainly point to two items that also appear in a Hive presentation slide deck.
- A 2024 study from the University of Chicago found that Hive’s AI detection tool outranked four other detectors in identifying AI-generated art.
- Hive also has a Pentagon contract for identifying deepfakes.
MIT Technology Review reported in December that Hive was selling its deepfake-detection technology to the US military. The new Cyber Crimes Center contract applies that broader AI detection capability to a child exploitation investigative context.
The filing’s logic is straightforward: synthetic CSAM is growing, investigators need to focus on real victims, and automated analysis may help manage the volume of digital content. The open question, based on the source, is how well general AI image detection performs when used on CSAM-related investigations.
The stakes of using AI against AI-generated abuse material
This trial shows a difficult turn in online safety. Generative AI is contributing to a surge in synthetic child abuse images, while AI detection is being tested as part of the response.
The potential benefit is clear. If investigators can more quickly distinguish AI-generated CSAM from material involving real children, they may be able to direct limited resources toward urgent cases. That is why the government filing emphasizes focusing investigative resources on cases involving real victims.
But the source also makes clear that this is still a test. The contract is for three months, its details are heavily redacted, and the effectiveness of such detection models in this specific setting was not addressed by the National Center for Missing and Exploited Children before publication.
For now, the Cyber Crimes Center’s contract with Hive AI is an experiment in using automated detection to manage a new form of digital evidence. Its significance lies in the problem it targets: a flood of AI-generated CSAM that can make it harder to identify where real-world abuse may be happening now.