AWS and Microsoft Exit Voice Biometrics as Google Pulls Back Amid Deepfake Risks

Amazon Web Services (AWS), Microsoft, and Google are scaling back or exiting their involvement in voice biometrics, signaling a major shift for a technology once promoted as a convenient authentication method. The changes highlight both technical limitations and growing concerns over security threats such as synthetic speech and deepfakes.

Vendors Step Back from Voice Authentication

AWS has announced that its Amazon Connect Voice ID product is being discontinued. New customers have been blocked since May 20, 2025, and full end of support is scheduled for May 20, 2026, according to company documentation. Microsoft will retire its Azure AI Speaker Recognition service on September 30, 2025, though Nuance Gatekeeper, another Microsoft-owned voice authentication product, remains available but appears to be de-emphasized. Reports also indicate that Google has removed public references and documentation for its Speaker ID technology, suggesting a step back from the market, though the company has not issued a formal announcement.

Analysts note that hyperscale cloud providers have faced persistent challenges achieving consistent accuracy for voice biometrics in real-world environments. Background noise, illness, and accent variations have all reduced reliability, while the rapid rise of synthetic speech technology has made systems more vulnerable to spoofing. Industry observers suggest these issues contributed to limited customer adoption and made voice authentication less attractive to major vendors. As one industry analysis put it, the hyperscalers have shifted resources to other areas where demand and performance are stronger.

Diverging Industry Strategies

The retreat by AWS, Microsoft, and Google contrasts with continued deployment in specific industries. In Australia, ANZ Bank continues to use its Voice ID feature to authorize mobile transactions over AU$1,000. The bank has positioned voice authentication as a convenient way to add security for large-value payments, underscoring that some financial institutions still see value in the technology despite broader industry retrenchment.

Meanwhile, security experts and researchers continue to highlight vulnerabilities in voice biometrics. Academic studies show that deepfake audio can bypass detection systems, raising questions about whether the technology can be trusted in high-security contexts such as banking or healthcare. Analysts suggest this risk has driven companies to prioritize more resilient methods such as multi-factor authentication and other biometric modalities, including facial and fingerprint recognition, which have demonstrated higher reliability in consumer and enterprise use cases.

The departure of major cloud vendors may slow standardization and innovation in voice biometrics, but smaller specialist firms continue to operate in the space. Providers such as Pindrop, Veridas, and Daon are still offering voice-based authentication solutions, though the future of large-scale deployments remains uncertain as the industry weighs convenience against evolving security risks.

Sources: Microsoft Learn, CX Today

–

By Ali Nassar-Smith