Topics we’re looking for.
SafeVoiceAgents welcomes work on — but not limited to — the following areas. In line with the exploratory nature of AAAI workshops, we especially encourage unpublished, ongoing, or emerging work.
Safety, alignment, and evaluation
Benchmarks, taxonomies, safety judges, value alignment, human evaluation, and red teaming for multi-turn voice agents.
Attacks, robustness, and guardrails
Jailbreaks, prompt injection, adversarial audio, refusal behavior, runtime monitoring, and safeguards for streaming or full-duplex interaction.
Privacy, identity, and misuse
Privacy leakage from audio; speaker identity exposure; consent and retention; spoofing, impersonation, provenance, and deepfake-enabled misuse.
Human factors, fairness, and high-stakes use
Trust, persuasion, overreliance, accessibility, and performance disparities across accents, age, disability, and speech impairment.
Deployment and governance
Auditing, incident reporting, release practices, data governance, and trustworthy deployment of voice agents.
The submission portal will be announced here. Check back closer to AAAI 2027.
We follow the AAAI author instructions and formatting guidelines.