Only 31% of HR leaders say they can spot a deepfake during a live interview, and fraud attempts have climbed over 2,000% in three years.

The fake candidate problem finally has a runtime answer

Video interview fraud and executive impersonation over video calls share the same blind spot: every existing security layer fails once a human believes what they are seeing. There is no firewall for a convincing face.

A background process that watches what you cannot

Scam.ai Announces Qualcomm Partnership, Launches Halo Deepfake Detection Model at Computex 2026 runs silently behind any video conferencing session and flags AI-generated or synthetic video in real time, with no footage sent to a cloud server and no change to how you already run calls. The model, called Halo, was built to run locally on Qualcomm-powered desktop hardware, which means detection happens on the device before the signal ever leaves the room. You open your normal call, Halo monitors in the background, and a flag surfaces when something synthetic is detected.

Executives and recruiters carry the most exposure here

• HR and talent teams conducting video interviews who need documented proof that the candidate on screen is real before an offer is extended
• CFOs and CEOs on high-stakes vendor or investor calls where identity fraud can trigger wire transfers or data disclosures worth millions
• Venture capitalists running founder due diligence calls where a spoofed founder could delay or corrupt a deal

Each of these roles takes frequent calls with unfamiliar faces, which is precisely the attack surface deepfake tooling was built to exploit.

Qualcomm’s device push just made on-device AI a security category

The Qualcomm partnership, announced at Computex 2026, gives Scam.ai hardware-level optimization access that cloud-dependent competitors cannot replicate at the device layer. If Qualcomm continues building on-device AI into its desktop ecosystem, real-time fraud detection could become a default expectation for enterprise hardware procurement.

What you can do with it starting now

• Run Halo passively during every recruiting video call to log synthetic video flags
• Protect executive video calls with zero changes to existing conferencing tools
• Detect AI-generated faces in real time before an action is taken on the call
• Keep all video data local, satisfying privacy requirements without a policy rewrite

Pricing not listed — check our directory.

Scam.ai currently targets desktop and is optimized for Qualcomm-powered devices, which limits deployment for teams running non-Qualcomm hardware at scale.

Pindrop and Sensity AI address synthetic media detection but focus on audio fraud and post-call forensics respectively, neither operates in real time on-device during a live call. Scam.ai is the first announced product to combine live detection with fully local processing at the desktop layer.

On-device AI is quietly becoming the new perimeter for enterprise security

The shift from cloud-based detection to local inference is accelerating faster than security teams have updated their vendor lists. We cover tools like this every Friday — subscribe here and we’ll send the best ones straight to you.