The Face in the Meeting Is No Longer Enough

The job interview looks ordinary. A candidate answers the question, the camera works, the voice matches the face. For a remote employer, those small comforts used to add up to identity. Generative AI has made the arithmetic less reliable. A convincing interview can conceal a different person, a synthetic face, or a voice that belongs to a model. GetReal Security has built its business around one unfashionable request: show your working before you trust the screen.

The company began as a meeting of two kinds of expertise. Hany Farid, a researcher known for digital media forensics, and Ted Schlein of Ballistic Ventures founded GetReal in 2022. Their premise was that the manipulation of media would become a security problem, rather than an oddity for fact-checkers. Ballistic incubated the company. In 2024, veteran cybersecurity executive Matt Moynahan became CEO. The team had to turn a forensic discipline into something an organization could use while a decision was still reversible.

The first crack appeared in the file

Early deepfake detection had a natural home: an image, audio clip or video file that someone could pause, upload and examine. GetReal Inspect, launched in 2025, is the company’s forensic workbench for that job. It looks for traces of editing and generation across multiple dimensions, then gives investigators a way to explain what they found. The method draws on machine learning, signal processing, traditional forensics and file analysis. A verdict without an explanation may be enough for a casual social feed. It is a harder sell in a boardroom, newsroom or legal dispute.

“AI alone isn’t enough to combat the malicious uses of deepfakes.”Hany Farid, co-founder and chief science officer

Farid’s line is also a description of the company’s operating model. Automation spots anomalies. Forensic tests probe them. In the difficult cases, a human investigator assesses context, provenance and what the content was meant to accomplish. GetReal Respond offers subscriber access to those specialists and forensic reports. GetReal Prepare covers readiness assessments, policy and training. The business, then, sells software and expert judgment together, with enterprise buyers directed to a sales conversation for pricing.

Then the suspect joined the call

A doctored file gives an investigator time. A live call does not. A finance employee may be asked to approve a payment while a familiar executive appears on screen. A help desk agent may be asked to reset credentials. A hiring manager may move a candidate through the final interview. The security decision happens in seconds, often inside a collaboration tool built for conversation, not evidence handling.

That is where GetReal Protect enters. The company expanded real-time coverage to Microsoft Teams and Cisco Webex in September 2025, and its May 2026 announcement said Protect works with Zoom and voice systems too. It analyzes face and voice signals during an interaction, flags suspicious identity changes, and can prompt the host or trigger policy actions. The company also describes an Identity Threat Graph that links appearances across meetings and affected users. A single suspicious participant becomes an incident with a history.

The distinction is timing. Conventional identity systems establish that someone signed in. GetReal’s newer product asks whether the apparent human remains credible as the conversation unfolds. Its software is meant to live inside existing work, which matters because staff cannot reasonably move every interview or payment request into a forensic laboratory. But a detection system is only as useful as the action it produces. That is why the product connects analysis to prompts, investigation and response, rather than stopping at a suspicious label.

A security budget meets a human habit

GetReal’s own research offers a revealing, if self-reported, measure of the anxiety. In a September 2025 survey of 668 IT, cyber, fraud and risk leaders at organizations with at least 1,000 employees, 41% said their company had hired and onboarded a fraudulent candidate. The survey cannot establish how many hires involved deepfakes, and it should not be read as a population-wide incidence rate. It does show why the hiring pipeline now belongs in the security conversation.

The same shift explains the investors. Forgepoint Capital led GetReal’s $17.5 million Series A in March 2025, with Ballistic Ventures, Evolution Equity Partners and K2 Access Fund among participants. Strategic investors included In-Q-Tel, Cisco Investments and Capital One Ventures. Their presence says less about a guaranteed winner than about the places where identity doubt now causes trouble: government, collaboration infrastructure and financial services.

GetReal’s partnership with Doppel is another clue. Doppel watches for brand and executive impersonation online; GetReal supplies content verification and forensic analysis. A fake executive video outside the firewall can become a legal, communications and security incident at once. The pairing broadens GetReal’s role beyond the meeting room without pretending every problem is solved by a meeting plugin.

What a buyer can copy before buying anything

The company’s most portable idea is a workflow, not a model. Identify the moments when a face or voice is treated as authorization: a candidate interview, a vendor bank-detail change, an account recovery call. Decide in advance which signals should pause the process, who receives the alert, what independent verification follows, and how evidence is preserved. A tool can accelerate that sequence, but it cannot invent the policy while a fraudulent transfer is already in motion.

There are limits. Deepfake detection changes as generation tools change; a result needs confidence, context and human review when the stakes are high. The product is built for organizations with collaboration and security workflows to integrate, so a small team seeking an occasional public image check may need less machinery. GetReal has not published a standard price or a public customer count. Those gaps matter in a purchase decision, even if they do not weaken the basic question the company asks.

GetReal’s name is almost impudent. In a world of synthetic faces, “real” sounds like a claim no software should make lightly. The more interesting promise is narrower: turn a vague feeling of trust into a decision supported by signals, evidence and a plan. The face in the meeting may still be genuine. Now someone has to check.