AI screening and GDPR when the vendor wants to check the candidate's face and IP address
Hirevue has launched fraud detection with face liveness checks, document verification and IP signals. The press release gives no error rates, and in the EU a face check is biometric data under Article 9 of the GDPR. The article sets out what employers need to settle before switching the checks on.
What happened
On 10 November 2026, Hirevue issued a press release through PR Newswire announcing new capabilities for detecting cheating and identity misrepresentation in hiring. The reasoning is that generative AI has made it easier to polish an application, generate convincing interview answers and impersonate another candidate. According to Hirevue, an application on its own can no longer confirm that a candidate has the skills claimed, or that the person evaluated is the person hired.
The platform could already flag browser switching, pasting from a language model and live use of one during an interview. New are four layers on top. Identity verification uses document checks and what Hirevue calls face liveness detection, a check that a living person is in front of the camera and matches the document. Location signals from IP address, VPN and proxy activity provide context when a candidate's location needs review. A checklist guides interviewers through verification steps. All signals are brought together in one report with evidence for human review and an auditable record.
Hirevue stresses that the employer chooses which signals are active and how much weight each carries, that candidates can be told about the process in advance, and that a flag prompts human review. A flag, the release says, does not by itself establish fraud or determine a hiring outcome.
What the press release does not say
There are no figures on accuracy. No false positive rate, no false negative rate, no independent validation, no price and no availability date. The release describes a product, not a result. This matters because a wrong flag lands on an honest candidate. The line about protecting authentic candidates only holds if the share of incorrect flags is known and low. A buyer has to ask for that.
A face check is biometric data under the GDPR
This is where the story becomes European. Article 4(14) of the GDPR defines biometric data as personal data resulting from specific technical processing of a person's physical, physiological or behavioural characteristics which allow or confirm the unique identification of that person. Facial images are named as an example. A face check whose purpose is to confirm that the candidate is the person on the document falls within that definition.
Article 9(1) prohibits, as a starting point, processing biometric data for the purpose of uniquely identifying a natural person. The prohibition is only lifted if an exception in Article 9(2) applies. The two most obvious are explicit consent from the candidate and processing necessary for obligations or rights under employment law, but the second requires a basis in EU law, national law or a collective agreement with appropriate safeguards.
Consent sounds like the easy route, but a candidate who wants the job is not in a free position when the employer asks. That is the question the data controller must answer before the check is switched on. On top of that, Article 35(3)(b) requires a data protection impact assessment when special categories of data are processed on a large scale. A face check of every applicant in a high-volume campaign is hard to call anything else.
IP address, VPN and proxy are personal data too
The location signals are less dramatic, but not free either. An IP address is personal data, and the data minimisation principle in Article 5(1)(c) limits processing to what is necessary for the purpose. Why do we need to know whether a candidate uses a VPN? If the answer is to check that the candidate is in the country the job requires, there is a justification. If the answer is that the signal happened to be available, there is not. The information duty in Article 13 also means the candidate must know that the signal is collected and why. Hirevue writes that employers can inform candidates in advance. In the EU, that is not a can.
Human review has to be real
A flag that only triggers human review is the right design and in line with Article 22 on automated individual decision-making. But the review only protects the candidate if the reviewer has the time, competence and mandate to overrule it. With a thousand applicants, the flag easily becomes the decision in practice.
Questions to ask before switching the checks on
Which legal basis under Article 9(2) covers the face check, and will it hold if the supervisory authority asks? Has the impact assessment been done, and who signed it off? Which false positive rates can the vendor document, and for which groups of candidates? Who reviews a flag, within what time, and can that person dismiss it? What is the candidate told, and how can they object? Which signals do we switch off because we have no justification for them?
Cheating with AI in recruitment is a real problem. But face and location checks move recruitment into the part of the GDPR where the starting point is a prohibition. That is not an argument against the checks, but for deciding with legal advice and documentation rather than a tick box in the vendor's settings.
Kilde: https://www.prnewswire.com/news-releases/hirevue-launches-advanced-fraud-detection-to-help-employers-validate-candidate-identity-and-skills-302933903.html