Tofu

What it is

Tofu sells candidate fraud detection that sits on top of your ATS and answers one question before a recruiter spends time on an applicant: is this a real person, applying from where they say, under their own identity? The core product is the Fraud Agent. It scores every applicant the moment they land in the ATS, checking email, phone, device, IP, LinkedIn, and GitHub against what Tofu says is 4B+ public data points and a “Fraudbase” built from 18M+ analyzed profiles. It then labels suspect profiles in the ATS and writes the evidence into the application notes. A second product covers the interview: deepfake and proxy-interviewer detection on Zoom, Google Meet, and Teams, plus AI note-takers. Around those two sit ATS Recon (a scan of the pipeline you already have), AI Resume Review, a Fraud API and Deepfake API for platforms, a Chrome extension, and a free tool that tells job seekers whether a recruiter who messaged them is real.

Tofu Technologies is run by co-founder and CEO Jason Zoltak, previously at the labor marketplace Shiftsmart. The company started as an AI resume screener and pivoted after customers reported that the voice on the interview didn’t match the resume they had screened. It announced a $5M seed on December 16, 2025, which Forbes reported was led by Slow Ventures, with Founder Collective, Night Capital, Ritual Capital, Wayfinder, and Garage.vc also listed on Tofu’s site. JumpCloud Ventures made Tofu its first investment on February 10, 2026, for an undisclosed amount. The same December day, Gem announced that Tofu powers Gem’s AI Fraud Detection Agent.

Why it matters now

• It is the engine inside Gem’s fraud agent. Gem’s agent reads six signal families: resume metadata, LinkedIn, email, identity, phone, and IP/device. It returns high, medium, or low risk with a stated rationale and claims 90%+ accuracy. Gem’s product page now describes the agent as available for Gem ATS only, with Greenhouse and Workday customers sent to sales for roadmap, and no longer names Tofu. If you run Greenhouse, Ashby, Lever, or Workday, buying Tofu directly gets you the same class of detection without moving ATS.
• Fraud volume on technical roles is no longer a rounding error. Checkr’s October 8, 2026 beta analysis flagged 36% of inbound resumes for review, 50% on software and AI engineering roles against 23–27% for sales, marketing, and ops. Tofu’s own numbers run higher: a customer testimonial says it flagged 46% of applicants on some tech roles, and the CEO cites 56%+ fake applicants in an average targeted pipeline. Treat the vendor numbers as upper bounds from targeted pipelines, not base rates.
• It covers the stage after the application. A clean application can still be followed by a proxy at the technical round. Tofu is one of the few vendors selling both the application score and live-interview detection under one contract, including on Google Meet, which several interview-detection vendors don’t cover.

Pricing reality

Tofu publishes its ladder, which most of this category doesn’t:

• Free Fraud Scan: $0, up to 10 applicants, once. No ATS connection needed. This is a trial, not a free tier.
• Growth: from $833/month, billed annually (about $10,000/year), covering up to 15,000 applicants and 500 interviews a year. It includes standard fraud detection on every applicant, risk scores written back to the ATS, a fraud dashboard, audit logs, and the Chrome extension.
• Enterprise: custom, for 15,000+ applicants, with unlimited applicant scans, advanced fraud detection, deepfake detection, ATS Recon, SSO/SCIM, and dedicated CS.
• APIs: Fraud API and Deepfake API are priced per call, quoted on volume.

At full Growth usage, that works out to roughly $0.67 per applicant. A team at 3,000 applicants a year pays about $3.33 each. The catch is that deepfake detection, ATS Recon, advanced verification, and AI Resume Review are all add-ons on Growth, and they aren’t priced on the page. The real-world band for a mid-size tech company buying application scoring plus interview deepfake detection is therefore above $10K/year. Get the add-on quote before you compare. Billing is annual only.

Best for

A TA leader at a 100–2,000-person company hiring remote engineers, IT, or finance staff through Greenhouse, Ashby, Lever, or Workday, at 2,000+ applicants a year, who wants fraud labels inside the ATS before the recruiter screen and doesn’t want to switch ATS to get them.

Do not buy it if you hire mostly in person for hourly roles, or make fewer than about 25 remote hires a year. A $4.99 identity check at offer covers that risk for far less.

Watch-outs

• Most flags at a low base rate land on legitimate candidates. VPN users, relocating candidates, and people with thin online footprints all trip signals. Zoltak himself argues against single-signal rules such as email age. Guard: route every flag to a human review queue with a second signal class required before action, and turn off any bulk-reject workflow until you have checked a month of flags against confirmed outcomes.
• A third-party fraud score used to reject candidates may raise FCRA questions. Tofu’s site shows SOC 2, GDPR, and Warden AI Assurance badges, but it doesn’t state an FCRA position. Guard: get Tofu’s FCRA position in writing during procurement. Have counsel decide whether a flag can ever be the stated reason for rejection, or only a trigger for an identity check.
• Interview deepfake detection analyzes faces and voices. For Illinois candidates that brings in BIPA and the AI Video Interview Act. Guard: add notice and consent to every interview invite on requisitions where detection runs, apply it to every candidate on the req, and put retention terms in the contract.
• It’s a seed-stage vendor in a category incumbents just entered. Guard: cap the first term at 12 months, confirm ATS label writeback works in your instance during the free scan, and keep the identity gate in a separate tool you could keep if Tofu goes away.

Alternatives, and when to pick them instead

• Greenhouse Real Talent. The native option for the largest mid-market ATS install base: IPQS-based fraud signals plus CLEAR identity verification, on the Plus or Pro tiers. Pick it when you already pay for those tiers. Buying Tofu on top only makes sense if you need live-interview detection.
• Checkr Fraud Detection Suite. Launched October 8, 2026 by the background-check vendor with 140,000+ customers. It covers resume fraud, image matching, deepfake detection, biometric ID, device, and location signals, from application through post-hire reverification. Pick it when Checkr already runs your background checks and you want one vendor and one adverse-action flow. Its pricing and ATS writeback weren’t published at launch, so test both.
• GetReal Security. The fastest-rising interview-detection entrant, a Gartner Market Shaper in 2026. Pick it when the CISO already funds it for help-desk and executive impersonation and interviews run on Teams, Webex, or Zoom.
• Persona. An identity gate rather than a fraud score. Pick it when you can fund only one control and want proof of who the person is at a named stage, from $250/month.

If none fit because you hire in person, you don’t have this problem to buy for. For where application scoring sits against the identity gate, the recorder, and the post-offer check, see the hiring integrity stack and AI interview fraud.