Reality Defender

What it is

Reality Defender is a New York deepfake-detection company that sells the same multimodal models through four products. RealMeeting puts a detector bot into live meetings and tells the host whether each participant’s face and voice look authentic or manipulated. RealAPI is the developer API and SDK set (Python, TypeScript, Go, Rust, Java, REST) for uploaded or recorded image, audio and video. RealScan is the drag-and-drop web app for the same files. RealCall scores voice in contact-centre and telephony traffic. The suite launched under those names on November 19, 2025.

For a recruiting team the question it answers is narrow: is the person on this interview a real face and a real voice, or a synthetic overlay? It does not tell you who the person is. That is the job of an identity check.

The company was founded in 2021 by CEO Ben Colman and came out of Y Combinator (W22). It expanded its Series A to $33M on October 22, 2024, led by Illuminate Financial with DCVC, Booz Allen Ventures, IBM Ventures and Accenture. KPMG LLP took an undisclosed minority stake on September 10, 2026 and says it will fold the detection into its cyber and fraud practices.

Why it shows up in hiring decisions now

• Gartner placed it as a Market Shaper in the Emerging Market Quadrant for Deepfake Detection: Startup Vendors, dated June 25, 2026. That is the same report that placed GetReal Security.
• The interview product is documented for Microsoft Teams. The RealMeeting user guide dated September 22, 2026 covers Teams only. An admin adds the user, a Microsoft Entra Global Administrator may need to grant consent, and the host adds the app from the meeting’s Apps menu. The bot then scans each participant in join order from a short video-and-audio sample. Results show only to the host, as Authentic or Manipulated with a confidence score, and nobody is removed from the call. A September 2025 demo ran the same pipeline as a Zoom bot, but the current product page only says “major enterprise meeting platforms.”
• The built-in alternative arrived. BrightHire put deepfake detection, powered by Zoom’s own models, into early access for Zoom Workplace customers in June 2026, with thinner coverage on Teams and Google Meet. If you run interviews on Zoom and already pay for BrightHire, you may not need a second detector.
• It is the only detection vendor in this group with a public price. Detection can be tested on your own recorded interviews before any sales call.

Pricing reality

The RealAPI page, checked November 6, 2026, lists three plans:

The page labels the $399 “Annual Billing” without saying whether it is the monthly rate. Third-party listings read it as $399 a month, about $4,800 a year, so confirm before budgeting. When you hit the scan cap, scans pause until the next cycle. RealMeeting is not on that page and is sold through Enterprise. The API also does not analyse live calls, only uploaded and recorded media.

At Business volume a scan costs about $0.40. For comparison, Persona starts at $250 a month for 500 identity verifications, and Checkr adds identity verification to a background report for $4.99. Our read: the Business plan is a cheap way to measure whether deepfakes show up in your interview recordings at all. A RealMeeting contract only makes sense once that measurement says they do.

Best for

A TA operations lead or recruiting-security partner at a company that hires remote engineers or IT staff on Microsoft Teams and wants evidence before buying. Start with the Business API on a sample of recorded interviews and move to RealMeeting only if real cases turn up. It also fits a team building fraud checks into its own interview tooling, because it is the one detector here you can call from code without a contract.

Skip it if your interviews run on Zoom with BrightHire, if you hire in person, or if you have no identity check at a named stage yet.

Watch-outs

• A “Manipulated” flag is a probability, and at low fraud rates most flags are wrong. The arithmetic is in the interview-fraud screening skill. Guard: write into the hiring policy that a flag routes the candidate to an identity re-check, never to a rejection, and keep the confidence score with the record.
• Platform coverage beyond Teams is not documented today. Guard: get the supported-platform list in writing, map every interview round to it, and make coverage of each round a pilot exit criterion.
• The bot is visible and scans face and voice. Participants see it join “similar to an AI notetaker.” Consent is an optional admin setting, and detection samples are kept unless an admin turns on “Delete files after analysis.” Illinois BIPA covers face geometry and voiceprints, and the AI Video Interview Act requires notice and consent. Guard: turn on the in-meeting consent prompt and file deletion before the first candidate call, and add notice to the interview invite.
• Vendor benchmarks are not your adversary. Detectors weaken on generators they were not trained on. Guard: test on your own known-good and known-bad clips during the Business-plan trial before you sign for RealMeeting.

Alternatives, and when to pick them instead

• An identity gate: Persona or CLEAR. Pick this when you can fund only one control. It proves who the person is, which a detector cannot. See AI interview fraud for why the gate comes first.
• GetReal Security. Pick it when the CISO already funds it for help-desk and executive impersonation, or when you need continuous identity checks across rounds on Teams, Webex and Zoom.
• Pindrop. Pick it when phone screens and voice are the main exposure or the contact centre already runs Pindrop.
• BrightHire on Zoom. The newest entrant, and the one that ships inside a tool many teams already own. Pick it when interviews are on Zoom and BrightHire is already your interview-intelligence layer.

If none fit because you hire in person, there is no detection problem to buy for. The hiring integrity stack shows where detection sits against application scoring, the identity gate and post-offer screening.