Scam.ai has announced a partnership with Qualcomm and launched Halo, an on-device deepfake detection model for live video calls, according to TechHQ. The company made both announcements at Computex 2026 in Taipei, where it appeared at Qualcomm's booth in the event's Agentic AI track.
According to TechHQ, Halo runs locally on desktop PCs, operates in the background during video conferencing sessions, and flags synthetic or AI-generated video in real time. Scam.ai said the Qualcomm partnership gives it access to device ecosystem resources and optimization support intended to help Halo run without cloud infrastructure. The product is described as optimized for Qualcomm-powered devices and available as of June 2026.
Halo Pushes Deepfake Detection to the Endpoint
The announcement places Scam.ai at the intersection of Enterprise AI, endpoint security, and live identity verification. Rather than analyzing uploaded media after the fact, Halo is being positioned as a continuous control for active meetings, interviews, and executive conversations.
That distinction matters. If the product performs as described, it shifts deepfake defense from a reactive investigation workflow to an in-session decision layer. For security and collaboration teams, that could make live video trustworthiness a procurement category of its own, alongside meeting platforms, identity controls, and fraud prevention systems.
TechHQ also reports that Scam.ai is targeting HR and recruiting teams, as well as high-value executives including CEOs, CFOs, and venture capitalists. The article cites two risk indicators: only 31% of HR leaders say they feel equipped to detect identity fraud in video interviews, and deepfake fraud attempts have risen by more than 2,000% over the past three years.
Why This Matters to Technology decision-makers
For IT, security, and infrastructure leaders, the most notable element is the on-device design. Local inference can reduce video data movement, avoid cloud processing dependencies, and potentially lower latency in real-time detection. That makes the product relevant not only to security buyers but also to teams responsible for endpoint strategy, compliance, and collaboration architecture.
The tradeoff is operational. Optimization for Qualcomm-powered desktops could deliver performance advantages on supported systems, but it may also complicate deployment in mixed fleets that include Intel, AMD, Apple, or other device profiles. Enterprises evaluating this class of Models should look beyond detection claims and ask practical questions about hardware support, background resource use, manageability, and user experience under load.
There is also a governance layer. Any tool that flags potential impersonation in interviews or executive calls will likely draw in legal, HR, compliance, and risk teams. False positives, retention of alert data, evidentiary standards, and escalation procedures will matter as much as model accuracy. Those considerations become sharper when the output could affect hiring decisions, financial approvals, or executive communications.
Qualcomm's Role Signals a Broader AI PC Security Play
Scam.ai's placement at Qualcomm's booth suggests that trust-and-safety use cases are becoming part of the AI PC narrative, not just productivity assistants or AI Agents. If vendors can show that edge inference improves both privacy and response time for fraud detection, AI-capable endpoints may increasingly be sold as security platforms as well as compute platforms.
That creates potential market pressure on cloud-based deepfake detection offerings, especially in enterprise settings where data handling and latency are purchasing factors. It also raises the bar for collaboration-security and identity-verification providers that do not yet address live video impersonation risks.
At the same time, buyers should treat the current disclosure as early-stage. TechHQ reports that additional enterprise integration details and platform partnerships will be announced in the coming months, which implies the broader deployment ecosystem is still taking shape. Scam.ai itself is described by TechHQ as a San Francisco-headquartered company focused on real-time deepfake detection for enterprise interactions, placing the launch squarely in the Startups and enterprise security conversation.
Sources and Methodology
This article is based on single-source reporting and should be read accordingly. The factual claims cited here come from TechHQ's June 29, 2026 report. Because no independent corroborating sources were provided in the source set, strategic conclusions about performance, compatibility breadth, and adoption should be treated as informed analysis rather than verified market consensus.
