Responsible AI at AuraSync

AI output is decision support, not a decision

AuraSync generates scores, summaries, integrity flags and recommendations with AI assistance. They are indicators, not measurements of a candidate's worth or ability, and they are not a medical, psychological or clinical assessment of any kind.

AI output can be wrong. It must not be the sole basis of a hiring decision, and customers are responsible for human oversight of any decision that affects a candidate.

Human oversight at every decision point

AuraSync is designed so that people, not models, make the consequential calls:

• ✦A person at the hiring organization reviews results and decides who moves forward.
• ✦Recruiters can open a candidate's underlying answers and materials at any time, and can disregard or override any score or recommendation.
• ✦Integrity flags are reviewed by a person, and the candidate is entitled to explain what happened.
• ✦Any candidate can ask for a human to review an AI-generated score or recommendation, and can challenge it.

Where AI is used in AuraSync

AI is used for specific, bounded tasks rather than open-ended judgment:

• ✦In the Assessment module: generating role-relevant questions, transcribing spoken answers, scoring answers, producing behavioral and psychometric indicators, and raising integrity flags.
• ✦In the ATS: resume parsing and drafting, interview question generation, job-description summarization and market benchmarking, where a customer has enabled those features.
• ✦In AskAura: answering hiring teams' questions about candidates and roles, grounded in the assessment evidence already in the platform.

Known limitations and sources of error

Every AI system has limitations, and we think it is more useful to name ours than to hide them.

• ✦Audio, video and device: results can be affected by audio and video quality, accent, connection quality and the device used.
• ✦Disability and assistive technology: results can also be affected by disability or by the assistive technology a candidate uses, which is why adjustments are available.
• ✦Generated text: AI-written summaries and answers can be imprecise or incomplete, and their exact wording can vary between generations.
• ✦Estimates, not ground truth: scores are derived from a structured assessment and are estimates; recruiters should read the underlying materials before relying on one.
• ✦Automation bias: a recommendation can invite over-reliance. We recommend treating it as one input and documenting the team's own reasoning.

Bias and fairness

No AI system is free from bias, and we do not describe AuraSync as bias-free. AuraSync aims to make hiring decisions more consistent and easier to review by attaching evidence to every result and keeping people in control of the decision.

Employment law differs by jurisdiction. Customers remain responsible for making sure their use of AuraSync complies with the laws that apply to them, and for monitoring their own hiring outcomes.

Integrity flags are signals, not findings

AuraSync's assessment integrity checks raise flags when something unusual happens during a session. A flag is an automated signal, not a finding of misconduct: ordinary things such as someone walking behind the candidate, poor lighting or a dropped connection can raise one. A person reviews flags, and the candidate is entitled to explain what happened. Genuine technical problems are not misconduct, and a candidate may ask to resume or retake.

Biometric processing

Where an assessment uses identity verification:

• ✦Facial verification uses a mathematical representation of a face, not a photograph, to confirm that the person taking the assessment is the person who was invited.
• ✦Taking part, camera monitoring and facial verification are three separate consents, each recorded with the version of the notice shown.
• ✦Facial verification data is not shared with the hiring organization.
• ✦Facial embeddings are deleted within 90 days of capture by an automated purge, and immediately if the candidate withdraws that consent.

Accessibility and adjustments

If any part of an assessment is difficult because of a disability or for any other reason, a candidate can ask for an adjustment, either through the hiring organization or by contacting us, and we will pass the request on and record it.

Candidate rights and data use

The Privacy Policy sets out candidate rights in full. In summary:

• ✦Candidates have the right to know that an application is being processed with AI assistance.
• ✦Candidates can ask for human review of any AI-generated score or recommendation and receive a copy of their assessment results.
• ✦AuraSync does not sell assessment data and does not use candidate data to train its own AI models.
• ✦AI providers act as processors under contract and may not use the data for their own purposes.