The Partnership on AI is established

Background

As AI technologies moved out of research settings and into products used by large numbers of people, questions about how such systems should be built, tested and fielded were being raised by researchers, policymakers and members of the public at the same time. The Partnership on AI was set up as a place where those questions could be worked on collectively rather than company by company.

Founding

The Partnership on AI states that it was formally established in late 2016 ^[1]. It describes itself as having been led by a group of AI researchers representing six of the world's largest technology companies: Apple, Amazon, DeepMind and Google, Facebook, IBM and Microsoft ^[1].

A second account, contained in Microsoft's written evidence to a UK parliamentary inquiry, records the formation as having taken place in September 2016, among Amazon, DeepMind/Google, Facebook, IBM and Microsoft, and adds that Apple joined later as a founding member ^[2]. The two records place the same event at slightly different moments, one with a month and one with the end of the year, and the later list of founding companies includes Apple.

Mission

The organisation's stated mission is to study and formulate best practices on AI technologies, advance public understanding of AI, and serve as an open platform for discussion about AI's influences on people and society ^[2].

That mission is broken down into stated goals. The first is to develop and share best-practice methods in the research, development, testing and fielding of AI technologies ^[1]. The organisation also aims to advance public understanding of AI, to provide an open and inclusive platform for discussion, and to foster aspirational AI work for socially benevolent applications ^[1]. In plain terms, the goals mix technical work — agreeing on how systems should be developed and evaluated — with public-facing work such as explaining the technology and hosting debate about its consequences.

An open platform, not a lobby

The Partnership states that it is not a lobbying organisation, though it intends to be a resource to policymakers where relevant ^[1]. The distinction matters for how the body is read: it presents itself as a place where practices are compared and written down, and where discussion happens, rather than as a vehicle for pressing governments on behalf of its member companies. Being a resource to policymakers is narrower than advocacy — it means the organisation expects its work to be useful to people writing rules, without campaigning for particular legislation.

Becoming multi-stakeholder

In 2017 the addition of six not-for-profit board members expanded the Partnership into a multi-stakeholder organisation ^[1]. The term describes a body in which parties other than the founding companies — non-profits, and by extension the research and civil-society groups they represent — sit in governance rather than only being consulted. For an initiative that began as a coming-together of large companies, that change altered who has a say in what the organisation studies and recommends.

Why it is worth recording

The founding is documented in two different kinds of source: the organisation's own account of itself, and a company's written submission to a parliamentary inquiry ^[2]. The second is useful because it fixes both a date and a founding roster in an official record, and because it shows the initiative entering policy discussion early, as evidence in a legislative examination of AI.

What the Partnership did not do at the outset was release a model or a product. It set out to define and circulate good practice, to explain AI to a wider public, and to keep a shared forum open — ambitions that are slower to observe than a launch, and whose value depends on whether the methods it promotes are actually taken up.