Deep Media, a Bay Area startup led by CEO Rijul Gupta, has attracted attention for its AI detection claims and military contracts, but skepticism persists due to its small and inexperienced team. The growing market for deepfake detection services raises concerns over accuracy and reliability, with experts questioning the effectiveness of such tools in practical scenarios.
Deep Media, a Bay Area-based startup, has gained significant attention for its claim of detecting AI-manipulated images, audio, and video with 99% accuracy. The company, led by CEO Rijul Gupta, has secured at least five military contracts totaling nearly $2 million since late 2022. This includes a $1.25 million contract to develop a detection tool for the U.S. Air Force to counter information warfare from Russia and China.
Despite its claims and contracts, skepticism surrounds Deep Media’s capabilities, primarily due to its small team composed mainly of relatively inexperienced members in the AI and deepfake detection field. The sole machine learning engineer at Deep Media graduated with an undergraduate degree in astrophysics two years ago, and the company reportedly lacks staff with specialized PhDs or forensic backgrounds.
The broader market for deepfake detection services is growing, with about 40 companies now offering similar services. Many claim high levels of accuracy, though the reliability of these tools has been questioned. Testing by various institutes suggests these tools might not be as effective in practical scenarios due to the evolving nature of AI-generated fakes.
Deep Media’s legitimacy was further put into question with its analysis of a video of Catherine, Princess of Wales, which other experts saw no evidence of AI manipulation. The startup has faced criticism for potentially overstating its detection capabilities without sufficient scientific backing.
The Pentagon and other institutions continue to seek robust solutions for deepfake detection, given the rising concerns over synthetic media affecting critical areas like national security and elections. However, the effectiveness and honesty of claims within this sector remain widely debated among experts.
