1. Information processed by AI
When you request a parse, ApplyOS sends the job description, any supplied source URL, and target-role preferences used for that request to OpenAI, along with instructions for extracting fields. The parser’s current request does not attach your uploaded resume file or send your entire workspace.
Job text can itself contain personal or confidential information. Remove anything you do not have permission to disclose before submitting it. The service does not need government identifiers, account passwords, or private medical information to parse a posting.
2. Stored records and provider handling
ApplyOS stores parser input, URLs, results, caches, status, and usage information as described in the Privacy Policy. These records can remain after an unsaved or failed parse. Operator access and account deletion limitations also apply to these records.
OpenAI’s published API data controls state that API data is not used for model training by default, unless the API customer opts in. OpenAI may retain inputs, outputs, and abuse-monitoring records under its endpoint rules and account settings. ApplyOS does not promise zero provider retention or that deleting an application immediately removes provider logs.
3. Human review and limitations
AI-generated fields may be wrong, incomplete, or inferred. Review compensation, currency and pay periods, location, employment type, experience requirements, recruiter information, work authorization, sponsorship, and deadlines before saving or acting on a result. A confidence value is a model assessment, not a guarantee of accuracy.
Target-role priority and relevance scores are organizational suggestions. They are not assessments of your employability, protected characteristics, eligibility, or hiring chances. ApplyOS does not use these scores to make employer hiring decisions or automatically submit an application.
4. Your choice and feature availability
You can create and edit applications manually without the AI parser. Access depends on your account’s enabled features and usage limits. Cached results may be reused for matching requests. Limits, models, and availability may change, and provider outages or rejected requests may prevent processing.