AI Transparency in LinkedIn's Hiring Solutions

A look into our AI development, data privacy and compliance practices for LinkedIn Recruiter and Jobs.

Our commitment to building a trustworthy platform

Powered by the world’s most dynamic talent network, AI at LinkedIn unlocks new ways for customers and members to connect to opportunity with clarity, confidence, and control.

Guided by our responsible AI Principles, we are committed to building AI systems that are trustworthy, safe, and human-centered.

These core Responsible AI Principles guide all our work at LinkedIn:

Advance Economic Opportunity
Uphold Trust
Promote Fairness and Inclusion
Provide Transparency
Embrace Accountability

Advance Economic Opportunity

Our products leverage a skills taxonomy to promote equal treatment—focusing on what members can do, not where they come from.

Uphold Trust

We take a proactive approach to privacy, security, and safety— ensuring our AI products earn trust through rigorous assessments, safeguards, and defense in depth strategies.

Promote Fairness and Inclusion

LinkedIn’s cross-functional Responsible AI team designs solutions and guardrails (e.g. our fair model analyzer and mitigation trainer) to ensure generative AI tools proactively address potential bias and discrimination.

Provide Transparency

We empower our customers with transparency and education, so they can make informed choices about how they engage with generative AI.

Embrace Accountability

We embed robust governance from the start– cross-disciplinary assessments and human oversight ensure our AI systems stay accountable, trustworthy, and aligned with our Responsible AI.

Learn more about our recent innovations and development practices

Visit our Engineering Blog

Enterprise grade safety and governance

Building trustworthy AI at LinkedIn means mindfully innovating. From rigorous model testing to strong governance and legal compliance, we strive to build reliable and responsible AI tools for our customers.

A look into our AI development, data privacy and compliance practices for LinkedIn Recruiter and Jobs.

Model Testing

LinkedIn's AI models undergo rigorous and recurring evaluations – including jailbreak, harmful or illegal content, hallucinations, stability, latency, and output quality – to track fulfillment failure rates and analyze user feedback signals to refine and enhance the AI experience.

Equal Treatment Testing

Responsible AI standards have been stood up at LinkedIn to address potential model bias. Any models used in our products are audited internally on a recurring basis and undergo testing to help ensure the security of our models in compliance with our policies.

Governance

Our comprehensive data access management policy includes a quarterly review of access rights to production environments based on roles, continuous monitoring for abnormal activity, an independent annual SOC 2 audit, and penetration testing.

Members Control their Data

LinkedIn members control their personal data to train generative AI models for content creation, anytime, right from LinkedIn. Additionally, we test our models and applications against potential unintended data extraction attacks.

AI Explainability: Hiring Assistant Spotlight

Hiring Assistant is the only AI agent for recruiters powered by the world’s most dynamic talent network. Built on trusted, real-time data from 1+ billion members, Hiring Assistant delivers high-quality shortlists and uncovers candidates you’d otherwise miss.

How do recruiters use Hiring Assistant?

Recruiters choose the roles they want to use Hiring Assistant for and specify the qualifications they’re seeking in candidates. Hiring Assistant aides by refining those qualifications into better search queries and retrieving and displaying candidates that match the recruiter’s chosen qualifications.

Illustration explaining LinkedIn hiring assistant

Keeps humans in control of decision making

Hiring Assistant does not make autonomous or automatic decisions. Hiring Assistant ensures humans remain central to the hiring process, mitigating excessive reliance on automation.

Complies with AI regulations

LinkedIn complies with applicable law in the jurisdictions where our products are offered. We strive to provide the resources that enable our customers to assess and ensure their compliance with laws and regulations that apply to them.

Does not process sensitive personal data

Hiring Assistant does not require the processing of sensitive personal data provided by customers. Additionally, it does not evaluate based on characteristics such as (but not limited to) name, age, gender identity, race, religion, sex, and disability. Hiring Assistant is powered by data from LinkedIn’s economic graph—information that members and companies share and update on LinkedIn.

Addresses bias, privacy and safety via testing and in-product safeguards

Hiring Assistant’s models were reviewed for equal treatment and harmful content- helping to identify, investigate, and mitigate any detectable biases. Additionally, any user inputs that prompt Hiring Assistant to return sensitive personal data will result in the user being blocked from proceeding until removed.

Sources, evaluates, and shortlists candidates based on your hiring criteria

Hiring Assistant transforms the recruiter’s hiring qualifications into several search queries, and evaluates candidate profiles to assess how well they align with the role’s qualifications to deliver a shortlist. Candidates are then ranked based on how closely they meet the recruiter’s hiring qualifications, as well as other interest signals.