DoD awards contract for deepfake detection to Hive

The U.S. Department of Defense has awarded a U$2.4 million contract for deepfake detection of video, image, and audio content to AI company Hive.

The two-year deal will reinforce the country’s national security and the agency’s readiness for the future spread of misinformation, threats, and conflict, according to the firm. Earlier this year, a fake audio message from a voice resembling the U.S. President Joe Biden reached thousands of people in New Hampshire, advising them not to vote in primary elections.

The San Francisco, California-based company will be used by the DoD’s Defense Innovation Unit (DIU) intelligence community to fight the risks of synthetic media and AI-generated disinformation.

“This prototype has the potential to enable the DoD to detect and counter AI deception at scale, maintaining our nation’s information advantage in an increasingly complex digital battlefield,” says Anthony Bustamante, DIU’s project manager and cyber warfare operator.

The tools will be deployed in an offline and on-premise environment. In September, Hive struck a deal with Nvidia to allow customers to run its AI models in their own environment through Nvidia NIM, a set of accelerated inference microservices that allow organizations to run AI models on Nvidia processors.

The tools can also be adapted for broader use, such as protecting civilian institutions from disinformation. Aside from deepfake detection, the firm will also provide liveness detection for images and video.

According to the company’s CEO and Co-Founder Kevin Guo, the AI model has been trained on both AI-generated and real content and can detect patterns and signals in AI-generated content that are not visible to the human eye.

“Turns out that every image generated by one of these generators has that sort of pattern in there if you know where to look for it,” Guo told MIT Technology Review.

The company was chosen among 36 firms. In November, the DoD also awarded a US$1.8 million generative AI defense contract to Jericho Security for developing cybersecurity solutions that can ward off phishing attacks for the Air Force.

Aside from deepfake detection, Hive also builds AI tools for content moderation for companies including Reddit, Zynga, Midjourney, Runway, BlueSky and Truth Social.

Article Topics

deepfake detection  |  Department of Defense  |  government purchasing  |  Hive  |  national security  |  U.S. Government