Audio Deepfake Detection and Voice Integrity for Trustworthy AI Systems

CandyVoice provides expertise in audio deepfake detection and voice integrity, addressing one of the most critical challenges for trustworthy AI and secure human–machine communication.

With the rapid adoption of generative AI, synthetic and manipulated voices pose increasing risks to identity, trust, safety and decision-making across sectors such as civil security, media, XR, telecoms and public services. CandyVoice brings deep technical and system-level knowledge to help projects understand, evaluate and mitigate these risks.

Our expertise covers:

• detection of synthetic, cloned or manipulated voices in real-time and near-real-time scenarios,

• assessment of voice integrity and authenticity under noisy, compressed or degraded conditions,

• design of voice trust layers combining detection, robustness and privacy-by-design principles,

• integration strategies for audio integrity within larger AI or multimodal systems,

• alignment with GDPR, AI Act and emerging European standards on trustworthy AI.

Rather than offering a single black-box solution, CandyVoice supports consortia, research teams and industrial partners with:

• technical guidance and architecture definition,

• contribution to use-case design and risk analysis,

• support for evaluation protocols, benchmarks and validation strategies,

• input to dissemination, standardisation and policy-oriented activities.

This expertise is particularly relevant for Horizon Europe projects addressing AI trust, security, misinformation, digital identity and human-centred AI, where voice is a critical but often underestimated modality.

------------
Audio deepfake detection
Voice integrity
Trustworthy AI
Synthetic speech
Secure communications
AI Act
GDPR
Misinformation
Identity protection
Multimodal AI
Horizon Europe