Skip to Content
OPINION

Notes from the Asia-Pacific region: China marks PIPL's 5th anniversary, Hong Kong clarifies AI requirements

China is easing some cross-border data transfers and strengthening AI oversight, while Hong Kong is clarifying privacy requirements for AI use.

Published

Contributors:

Barbara Li

CIPP/E

Partner

Reed Smith

Editor's note

The IAPP is policy neutral. We publish contributed opinion pieces to enable our members to hear a broad spectrum of views in our domains. 

On 1 Nov., China's Personal Information Protection Law marked its fifth anniversary. In five years, China has built a personal information regime that sits alongside the Cybersecurity Law and the Data Security Law, supported by implementing rules, national standards and a cross-border transfer framework. 

Enforcement has matured in parallel. Regulators have moved from one-off campaigns to a continuing cycle of testing, publication, re-testing and removal, and prosecutors now combine criminal liability with public-interest litigation. The Supreme People's Procuratorate marked the anniversary by publishing model cases.

Local negative lists continue to ease cross-border transfers. Data that falls outside a free trade zone negative list may be exported from China without a security assessment, standard contract or certification. On 11 Nov., the Anhui Province branch of the Cyberspace Administration of China and other provincial authorities jointly issued the Anhui FTZ data export negative list, 2026 edition, and its implementing measures.

Two days earlier, Zhejiang replaced its 2024 list with a 2026 edition. The new edition retains the B2B e-commerce and clearing and settlement sectors and adds 11 bulk-commodity entries covering seven scenarios and 38 data items. Zhejiang has also streamlined its filing process to two steps: a single filing, after which data export may proceed.

Alongside the established tech and digital hubs of Beijing, Shanghai, Shenzhen and Guangzhou, Zhejiang and Anhui are major investment destinations with fast-growing digital economies. Multinationals should check the new lists against the transfer scenarios of each group entity registered in these zones and consider how to take advantage of these favorable local rules.

On 24 Oct., China's Ministry of Industry and Information Technology issued a notice on network access for AI-agent terminal devices. Devices with pre-installed AI agents must undergo a risk assessment before network access, and agent developers must strengthen adaptation verification and data security management. Most significantly, over-the-air upgrades are now subject to expert review. Businesses can no longer assume that software updates pushed after a product is on the market fall outside the network access regime: an update that changes an agent's functions, permissions or data processing may require review before release.

On enforcement, the CAC reported 6 Nov. that platforms had removed more than 47,000 items and dealt with 3,961 accounts that used AI to impersonate public figures. Under the 2026 personal information protection campaigns, testing bodies named 119 apps, mini-programs and software development kits, covering transport, hospital and school services. The primary wrong-doings include missing consent, incomplete disclosure of third-party SDKs, ineffective mechanisms for exercising user rights and inadequate protection of children's data.

In Hong Kong, the Office of the Privacy Commissioner for Personal Data issued the Artificial Intelligence: Model Personal Data Protection Framework 28 Oct., together with a checklist for employees using generative AI. Its central position is that uploading Hong Kong personal data to an AI model hosted outside the country is a cross-border transfer under the Personal Data (Privacy) Ordinance and is not lawful by default. 

The guidance also confirms three further points: identifiable data that is publicly available still requires a lawful basis before it is used for training; AI outputs, including hallucinations, must be monitored for leakage of personal data; and removing names and telephone numbers alone does not make data anonymous. Organizations using Hong Kong as a regional hub or preparing for a Hong Kong listing should map their generative AI tools by server location, update their privacy notices and keep records of AI data flows ready for due diligence.

This article originally appeared in the Asia-Pacific Dashboard Digest, a free weekly IAPP newsletter. Subscriptions to this and other IAPP newsletters can be found here.

This content is eligible for Continuing Professional Education credits. Please self-submit according to CPE policy guidelines.

Word count: 574

CPE credit badge

Contributors:

Barbara Li

CIPP/E

Partner

Reed Smith