Diopter VS Code Parcel

Diopter's answer

Diopter scores the live call, not a file after the fact. Most detection tools ask you to upload media and wait for a verdict. Diopter runs alongside a Zoom, Microsoft Teams, Google Meet or Webex call and scores every participant continuously for the full length of the call, including someone who is not speaking.

It also combines four signals rather than one. Video deepfake detection, audio deepfake detection, identity and payment verification, and policy alignment resolve into a single recommended action, so a security or finance team gets one decision rather than four dashboards.

The verdicts are deliberately honest. Bands are AI, mixed, clean and inconclusive, each with a confidence level rather than a flat yes or no. If a call could not be fully checked, Diopter reports that rather than defaulting to clean. It does not identify who is speaking, uses no voiceprints or enrollment, and never treats accent, region or language as a signal.

Diopter's answer

Because the attack happens on the call, and that is where Diopter works.

Identity verification tools check a person once, usually at onboarding or account recovery. That does nothing for a wire approval call three months later where the person on screen is a deepfake. Single-frame or file-based detectors judge a moment of media rather than the conversation shaping the ask.

Diopter runs for the whole call and scores every participant continuously, on Zoom, Microsoft Teams, Google Meet and Webex, and on VoIP and conference phone calls through Zoom Phone, Teams Phone, RingCentral and Dialpad. It combines synthetic media detection with identity and payment verification and policy alignment, then returns one recommended action.

It is also straightforward to deploy and clear about its limits. No caller-side install, works with a meeting bot or bot-free, on-prem and hybrid supported, configurable retention including zero data retention, and rollout through existing MDM such as Intune and Jamf. SOC 2 Type II in progress.

Diopter's answer

Security and fraud teams at organisations where a phone or video call can move money, grant access or hire someone.

The people who buy and run it are CISOs and security operations, fraud operations, and IT. The people it protects usually sit elsewhere in the business: finance and treasury staff approving wires, accounts payable teams handling vendor payment changes, recruiters and hiring managers interviewing candidates, executive assistants fielding urgent requests, and IT help desk agents resetting credentials.

The common thread is a live conversation where someone is asking for something consequential and the only proof of who they are is their face or their voice.

Diopter's answer

Diopter was founded in April 2026 by Rohan Gupta and Jeremy Pippin, after they watched generative AI move from productivity use cases into impersonation, fraud and social engineering.

Both came from QuillBot, where Rohan was founder and CEO of a generative AI company with more than 35 million monthly users, and Jeremy was VP of Product, having previously held the same role at FanDuel. They had spent years on the building side of generative AI and could see how cheap and scalable the attack had become while defence stayed expensive and manual.

The belief that shaped the product is that fraud is rarely detectable from a single frame or voice sample. It unfolds across a conversation, so the arc is the unit of analysis: authority, urgency, isolation, escalation and the ask, scored continuously through a live call.

The company is based in Chicago, Illinois.