China’s top legislature is advancing a new draft regulation — the Algorithmic Employment Decision-Making Management Rules — that for the first time assigns legal liability when AI systems make hiring, promotion, or termination decisions. The draft, circulated for public comment in June 2026, classifies AI-driven HR tools into three risk tiers and requires foreign employers using them in China to maintain human override capability and audit trails. If your China operation uses any automated screening, video-interview scoring, or performance-rating tool, this regulation will directly affect your compliance obligations starting in 2027.
Why It Matters
AI-powered HR tools have flooded the China market. By mid-2026, an estimated 38% of companies with more than 500 employees in China use some form of algorithmic hiring — from resume-ranking systems to AI-scored video interviews. Platforms like Moka, Beisen, and Zhaopin’s AI module process over 4 million job applications per month through automated screening. The efficiency gains are real: companies report a 40–60% reduction in time-to-hire when using AI pre-screening.
But the liability question has been a legal vacuum — until now. In 2025, a Shanghai court ordered a tech company to pay RMB 180,000 in damages after its AI hiring tool systematically downgraded female applicants for engineering roles. The court ruled the employer liable even though a third-party vendor provided the tool. The 2026 draft regulation codifies this principle: the employer is always the responsible party, regardless of who built the algorithm.
This matters for foreign companies in particular because China’s Personal Information Protection Law (PIPL) already imposes strict requirements on automated decision-making that produces “significant impact” on individuals. The new rules layer employment-specific obligations on top of PIPL, creating a dual-compliance burden that doesn’t exist in most home jurisdictions. As China Briefing has reported, this regulatory layering is becoming the norm across China’s tech governance landscape.
The Details
The draft regulation creates three risk tiers. Tier 1 covers basic admin tools — scheduling, payroll calculation, leave management. These require only a public disclosure notice and an opt-out mechanism. Tier 2 covers screening and assessment tools — resume ranking, skills testing, video interview scoring. These require a mandatory human review step before any adverse decision, plus an annual third-party bias audit.
Tier 3 is the high-risk category: tools that make or substantially influence hiring, promotion, demotion, or termination decisions. These require all Tier 2 obligations plus: real-time explainability (the system must generate a plain-language reason for each recommendation), data localization (all processing must occur on servers in mainland China), and a dedicated Algorithm Ethics Officer — a named individual with sign-off authority who can be held personally liable for non-compliance.
For foreign companies, the data localization requirement is the biggest operational hurdle. If you use a global HR platform (Workday, SAP SuccessFactors, Oracle HCM) with AI modules hosted outside China, you’ll need to either migrate the China instance to a local server or disable the AI features for China-based employees. The cost of a dedicated China-hosted instance typically runs $15,000–$40,000 per year for a mid-sized operation of 50–200 employees.
The draft also introduces a whistleblower provision: any employee can request a human review of an AI-driven decision within 30 days, and the employer must respond within 15 working days with a written explanation. Failure to respond triggers a rebuttable presumption of discrimination in any subsequent labor arbitration — a significant shift in the burden of proof.
What You Should Do
The regulation is expected to take effect in mid-2027, but the preparation window is now. Here’s what to do in the next 90 days:
- Inventory every AI tool touching your China workforce. This includes recruiting platforms, onboarding systems, performance management software, and even LinkedIn Recruiter’s AI-matching features if used for China-based roles. Document the vendor, the specific AI features used, and where the data is processed.
- Check your vendor contracts. Most HR SaaS agreements allocate all compliance liability to the customer. The draft regulation makes this non-delegable — you can’t contract out of it. Negotiate audit rights and data processing location guarantees in your next renewal cycle.
- Map a data localization plan. If you’re using a global HRIS with AI features processing China employee data, start the migration conversation now. The typical timeline for a China-instance migration is 4–6 months.
- Identify your Algorithm Ethics Officer. This doesn’t need to be a new hire — your China HR director or compliance lead can take the role — but they need documented training and explicit sign-off authority.
One Data Point
The number to remember: 38% — the share of large employers in China already using algorithmic hiring tools, a figure that has doubled since 2023. If you’re not using AI in HR yet, your competitors almost certainly are. The regulation isn’t about stopping AI adoption — it’s about making sure you can defend the decisions it makes.
Where to Go From Here
HR compliance in China sits at the intersection of labor law, data privacy, and corporate governance — each with its own regulatory body and enforcement regime. For the data privacy side of the equation, see our PIPL Compliance Guide for Foreign Employers. If you’re building a China team from scratch, our China Hiring and Labor Law Primer covers employment contracts, social insurance, and termination rules.
— China Gateway 360 —
Remote China market entry support, built around execution.
