Measures for the Administration of Data Security of Banking and Insurance Institutions, artificial intelligence and automated decisions
In force since .
An AI sector rules rule binding private bodies.
- Instrument type
- a regulation made under an act
- Obligation class
- Governance, Disclosure
- Audit expectation
- none
What it requires
- Applies to banks, trust, finance, leasing and consumer finance companies, wealth management companies, insurers, insurance asset managers and insurance groups established in China that use automated decision analysis or artificial intelligence in their business (Articles 2, 50 and 52); you are reached by declaring that you are a bank, a lender or an insurer.
- Keep data processing transparent and results fair and reasonable in automated decision analysis, model algorithm development and data labelling (Article 50).
- Manage artificial intelligence model development and use centrally, set an admission mechanism for models and algorithm products you bring in from outside, and keep models verifiable, auditable and traceable (Article 50).
- Before an information system or model algorithm goes into use, review the reasonableness, legitimacy and explainability of the data and model use and its effect on the people concerned (Article 51).
- Explain and disclose how data affects a decision result, monitor automated processing in real time, and prepare risk mitigation including an alternative plan to withdraw from the artificial intelligence application, emergency plans and drills (Article 52).
- Take effective measures to protect individuals' lawful rights in algorithm design, training data selection and model generation, and keep automated decisions that use personal information transparent and their results fair and just (Article 62).
These lines are LexLint's reading of the law's text.
Who enforces it
Settledness
- As of
What record it asks for
Logging duty
Articles 50 to 52 and 62 require no log or register; the data operation log duty sits in Article 43 and is held in the security document.
- Kind
- None
- As of
What this law does
When a banking or insurance institution carries out automated decision analysis, algorithm model development or data labelling it must keep data processing transparent and results fair and reasonable. It must manage the development and use of artificial intelligence models centrally, set an admission mechanism for externally sourced model algorithm products, and keep models verifiable, auditable and traceable.
Before an information system or model algorithm goes into use, the institution must carry out a data security review of the reasonableness, legitimacy and explainability of the data and model use.
An institution that uses artificial intelligence in its business must explain and disclose the effect of data on decision results, monitor automated processing and system results in real time, and prepare risk mitigation including an alternative plan to withdraw from the artificial intelligence application.
When it designs algorithms, selects training data and generates models the institution must take effective measures to protect individuals' lawful rights, and automated decisions that use personal information must be transparent and their results fair and just.
When LexLint raises it
When your app profile says your app is a bank, is a lender or is an insurer.
Back to the example · Lint your app
Page updated .
cn-50-52-62-measures-administration-data-securit