2026.09: Expanded candidate search filtering, rules-based match narratives, translation reliability improvements
AdeptID's September 2026 release adds new filtering options to Candidate Search, updates how Match Narratives are constructed, moves translation to more reliable model infrastructure, and improves the reliability of candidate feed ingestion.
Expanded Candidate Search filtering
What changed
New not_one_of exclusion operator: All custom filters that support one_of (string, integer, float, and date filters) now also support not_one_of. This returns only candidates whose value matches none of the supplied values.
New candidate_employers filter: Candidate Search now supports filtering on a candidate's employer(s), with the option to filter on either the current/primary employer or all known employers. The filter accepts an array of employer names and supports both one_of (include) and not_one_of (exclude).
Example usage: Exclude a hiring employer's current staff by combining candidate_employers with not_one_of.
Endpoints impacted
Existing
Action required
None. To use the new filters, add candidate_employers and/or not_one_of to your Candidate Search requests.
Rules-based match narratives
What changed
Narratives constructed with rules, not LLMs: Match Narratives are now built entirely from a series of rules and logic applied to the quantitative evaluations from AdeptID's audited models. No LLM is used to derive any assessment of fit between candidate and job. This makes narratives more consistent, reduces latency, keeps them aligned with match scores, and further mitigates the risk of bias.
Endpoints impacted
Existing
Action required
None. Request and response formats are unchanged.
Translation reliability improvements
What changed
More reliable translation models: Candidate and job translation have been migrated to more reliable models with greater availability and improved translation quality. Customer and candidate data continues to be processed by AdeptID under the same data privacy, data retention, and contractual commitments as before, with no impact to fairness/bias during matching.
Endpoints impacted
Existing
Action required
None.
Candidate feed ingestion reliability
What changed
Partial success for batch upserts: Previously, a single candidate profile that couldn't be stored could cause an entire upsert batch to fail. Now the valid candidates in the batch are stored, and any failing candidates are reported individually in the existing errors response.
Better handling of unusual characters: Profiles containing unsupported characters in free text, such as null characters copied from resumes, no longer cause ingestion errors.
Endpoints impacted
Existing
Action required
None.
AdeptID products are designed to support hiring and search (candidate sourcing, job search) decisions, not automate them. AdeptID systems are deployed within talent matching workflows where human judgment remains central to final outcomes. AdeptID believes that AI, especially when used in high-risk applications, must exceed high expectations for transparency, fairness, and accountability. This release has passed AdeptID's series of tests and controls for accuracy and fairness. See https://trust.adept-id.com/ for more information.