How we engaged with a leading Chinese technology firm.
Background
Founded in 2000 as a search engine, Baidu adopted artificial intelligence in 2010 to enhance content discovery and is now a leading Chinese AI company, with products and services that reach over a billion devices monthly. We have been engaging with Baidu on AI and digital rights since 2019, when we first encouraged the company to establish and publish AI governance principles, as we believe that doing so helps reassure investors that a company is appropriately mitigating risks associated with AI development and deployment.
Our engagement
We have been engaging with Baidu since 2013 on a range of topics, including board effectiveness, business purpose and digital rights. In 2019, while attending an autonomous driving trial at the company’s headquarters, we raised our concerns over AI ethics related to autonomous driving. We asked the company to develop and publish principles on how it uses and develops artificial intelligence, and how it addresses any ethical concerns it identifies. In 2020, in a letter to the chair and CEO, we outlined our concerns regarding AI and data governance. The company subsequently provided us with a detailed response, explaining its processes for blocking harmful content, training customers on AI use, and its approach to collecting user feedback and reducing algorithmic bias.
We shared our newly published EOS Digital Rights Principles in 2022 and offered to have a face-to-face meeting at the company’s offices that year to discuss our expectations regarding AI governance principles. Having not yet met, we made further requests to meet with the company in 2023. We reiterated our expectation that the company should publish AI governance principles.
We wrote to investor relations after the company’s third-quarter 2023 earnings call, which highlighted the wide range of AI use cases for Baidu’s advertising and cloud business partners. We explained that while we welcomed the recent establishment of a technology ethics committee, it would provide additional reassurance to investors regarding AI risk mitigation if the company published comprehensive AI ethical use principles.
In 2024, Baidu agreed to an in-person meeting at its head office in Beijing. We had a constructive meeting in which we welcomed Baidu’s recent publication of its AI governance principles termed “Baidu AI Ethics Measures.” We suggested that it should consider incorporating a description of “no-go” areas, or AI use cases that would create an unacceptable level of risk. We discussed several of the company’s initiatives through which it helps employees in different sectors of the economy to upskill with AI. We asked the company to include examples of these initiatives in future reporting.
Changes at the company
In 2019, the Beijing Academy of Artificial Intelligence published ethical AI standards backed by several technology firms including Baidu. In 2020, the company reaffirmed its active participation in the global initiative, Partnership on AI (PAI), and said that it was working with partners under these initiatives on case studies to demonstrate its efforts in AI fairness and anti-bias.
In its 2020 ESG report, Baidu announced four principles of AI ethics, proposed by its chair and CEO, Robin Li. These are: safety and controllability (the highest principle); equality (promote more equal access to technologies and abilities); empowerment (empower mankind to learn and grow, rather than surpassing and replacing humans); and freedom (bring more freedom and possibilities to humankind). In November 2023, Baidu established a technology ethics committee. The company then published AI governance principles in August 2024, which build upon the four principles announced in 2020. The principles cover many of the key aspects that we would expect to see, including core principles, oversight by the technology ethics committee, ongoing AI training, participation in the development of industry standards, and stakeholder engagement.
