Product Launches
Neural Defend's Arcas Models Take First and Second on Podonos Audio Deepfake Leaderboard
By News Desk ·
Key takeaways
- Neural Defend's Arcas 290M and Arcas 90M models placed first and second among 28 detection systems on Podonos's public audio deepfake benchmark, each scoring 99.93% on 4,524 clips.
- The smaller Arcas 90M runs at 89 ms per file, a quarter faster than its larger counterpart, and operates offline on a phone.
- Pindrop ranked 13th at 95.05%, Hive 17th at 83.53% and Reality Defender 18th at 71.27% on the same leaderboard.
Neural Defend's Arcas 290M and Arcas 90M models placed first and second among 28 detection systems on an independent, publicly scored audio deepfake benchmark run by Podonos, each scoring 99.93% on 4,524 audio clips. ANI News published the ranking on November 12, 2026.
How the Arcas Models Scored on Podonos
Each model made three errors across the 4,524 files, with a false-positive rate of 0.1% and a false-negative rate of 0.0%. Podonos described the result as "the closest any system has come to a clean sweep of this set." Of the 28 systems on the leaderboard, 18 are commercial products from 16 vendors, and the remainder are open-source and research models.
The smaller Arcas 90M matches the top score and runs a quarter faster at 89 ms per file. It is small enough to run on a phone without an internet connection, so audio never leaves the handset or the customer's network. Arcas 290M runs in the cloud through an API or SDK, on-premises, or in a cloud of the customer's choosing.
"We sent Podonos our predictions without knowing the answers, and we came out first," said Sivashankar Selvarajan, CTO of Neural Defend. "Every vendor claims 99%. This is the first time the whole market has been scored on the same files by someone with no stake in the result."
Neural Defend is deploying the models in banking and payments, call centres, KYC, insurance, ride and delivery platforms, matrimony and dating, and media. "A cloned voice now costs almost nothing to make; we intend to make it cost just as little to catch," said Piyush Verma, CEO of Neural Defend.
Neural Defend's Prior Zee News Verification System
According to akta.pro company records, Neural Defend partnered with Zee News (ZMedia) in October 2025 to launch an AI-powered deepfake verification system for news media, which lets audiences upload suspicious video, audio or image content for authenticity analysis within seconds. The same records show a March 2025 pre-seed round of $600,000 led by Inflection Point Ventures, with participation from MIT SBXI, Techstars and Soonicorn Ventures, earmarked for research and development and geographic expansion.
Pindrop, Hive and Reality Defender on the Same Board
Pindrop, which raised $100 million in financing in 2024, ranked 13th at 95.05%. Hive, valued at $2 billion and holding a US Department of Defense deepfake-detection contract, ranked 17th at 83.53%. Reality Defender, now supplying Orange Business, ranked 18th at 71.27%.
Stream Engine's Air Force Laboratory Delivery
Stream Engine delivered a containerized deepfake detection analytic to the Air Force Research Laboratory's Information Directorate under a CRADA, reaching test-ready status in June 2026, PR Newswire reported on November 3, 2026. AFRL recommended continuing the collaboration, with formal test and evaluation as the next phase.