Analysis of the First Revision of the Global Digital Compact: Implications for AI Governance

· 8 min read
By , , and

Following the release of the Zero Draft of the Global Digital Compact (GDC) on April 1, 2024, the co-facilitators of the process released a first revision (Rev. 1) on May 15, 2024, based on Member State consultations and input. Below, we highlight the key changes made to Objective 5 (the section focused on the governance of AI and emerging technologies), discuss what these changes imply, and outline why we think they’ve been made.

Overall, Rev.1 of the GDC remains a reasonably high ambition text, in that it preserves almost all of the Zero Draft’s institutional proposals in objective 5 (despite the pushback from Member States during the first reading over potential duplications with existing mechanisms). The co-facilitators have also simplified the structure of objective 5 by deleting the sub-sub-sections and removing some detail from the institutional proposals. 

In sum, key changes made to Objective 5 include:

  • A weakening of the overall ambition regarding the UN’s role in governing AI and emerging technologies;
  • The addition of a new paragraph outlining key governance functions, inspired by the High-Level Advisory Body on AI’s interim report;
  • A modification of the proposed ‘Annual Global Dialogue on AI Governance’ to a Contact Group on AI governance;
  • A broadening of the mandate of the proposed International Scientific Panel on AI to include emerging technologies;
  • A simplification and clarification of roles for the GDC’s follow-up and review processes.

Key changes to note in objective 5 + follow-up and review:

1. Weakening of overall ambition of the UN with regard to objective 5

What? Changes to the language indicate a slightly weakened ambition, or at least a downgrade to the centrality of the UN’s role in the governance of AI and emerging technologies. For example, the title of objective 5 has become ‘Enhance international governance of…’ instead of ‘Govern…’, and paragraph 52 now refers to the GDC as a unique opportunity to ‘advance’ international governance of AI and emerging technologies, rather than ‘establish the foundations for’ it. In paragraph 51, the role of the UN is now recognized as ‘vital’ but no longer ‘central’.

Why? This likely reflects the position of some Member States that the UN should not be leading international governance of AI and emerging technology, but only playing a complementary role, amongst and alongside other initiatives.

2. Addition of a new paragraph outlining key governance functions 

What? The co-facilitators have added a new paragraph (para. 52) that outlines 3 objectives for the GDC to “advance international governance of AI and emerging technologies”:

  • assess future directions and implications and promote scientific consensus, 
  • support compatibility, and interoperable norms, and 
  • build capacities to access, develop, use and govern AI and emerging technologies for sustainable development.

This clause echoes AI governance functions 1-5 outlined in the interim report of the High-Level Advisory Body on AI. However, Rev.1 still contains no mention of the two remaining functions identified by the HLAB on AI (functions 6 and 7), namely monitoring and reporting of risks/incidents, and compliance and accountability.

Why? This addition seems to provide a clarity of purpose that was missing in the Zero Draft, which skipped straight to the proposal of new institutions without first outlining the governance gaps to be filled. Rev.1 instead outlines what Member States want to do (para. 52), and then how they commit to doing it (paras. 53-61). This may facilitate understanding of the overall approach and also agreement by Member States, by allowing them to negotiate the functions separately from the proposed mechanisms.

3. Modification of the proposal for an ‘Annual global dialogue on AI governance’

What? Rev.1 contains a proposal for an International Contact Group on AI governance (para. 53b), which replaces the Zero Draft proposal for an annual global dialogue. This Contact Group is proposed as intergovernmental, rather than multistakeholder, in nature, but its stated purpose remains similar to that of the dialogue proposed in the Zero Draft - that is, to build shared understandings on governance, risk management and safety frameworks and to promote interoperability across governance approaches. Although Rev.1 mentions that this Contact Group’s work should be “supported by multistakeholder engagement on AI governance including the annual AI for Good Summit”, this revised proposal is nevertheless a clear shift towards prioritizing intergovernmental discussions on AI governance.

Why? The changes in Rev.1 could be the co-facilitators’ way of clarifying that the intent is to create something fundamentally different from existing efforts and that this function is not something that should be subsumed by the Internet Governance Forum or any other existing body. 

Note: The choice of a “Contact Group” to fulfill this function is interesting, given that such groups are traditionally informal groups that are formed to help manage peace and security crises (although there are some recent examples that go beyond this domain).

4. Broadening of the mandate of the international Scientific Panel

What? In the Zero Draft, the Panel was on AI only, but in Rev.1, it has become the International Scientific Panel on AI and Emerging Technology. Some of the detail of what it will do has been removed. For example, Rev.1 states that the Panel will issue reports, but the six-monthly reporting cycle is gone. Also gone is the mention of providing ‘early warning’. Meanwhile the new task of contributing to ‘mitigation measures’ has been added.

Why? The addition of ‘emerging technologies’ could reflect a wish for the panel to be more forward looking and to adapt its work and relevance to future technological developments. This sentiment is reflected elsewhere in Rev.1 which, in several places, seems to have broadened the focus from AI only to AI and emerging technologies (e.g. in para. 52). The diminished level of detail in terms of what the Panel will do likely aims to make it easier for Member States to agree on the fundamentals only (ie. that this is a governance gap to be filled), leaving the modalities and terms of reference to be proposed later on. This could provide more time for Member States to think through what institutional design will be most fit-for-purpose.

5. Simplification and clearer roles for GDC follow-up and review

What? The final section of Rev.1 contains a few changes that are relevant to the governance of AI and emerging technologies, as well as more broadly. 

For one thing, there is a subtly increased insistence on strengthening / leveraging multistakeholder platforms (additions in paras. 63 and 66) and the WSIS Forum in particular. 

There is also a noticeable effort to simplify the follow-up procedures, with the deletion of a reference to a new reporting portal for Member States (para. 72) and the proposal for a high-level meeting  to review the GDC seemingly becoming a one-off meeting, instead of taking place every two-years.

The new language also provides greater clarity in terms of the envisaged roles of WSIS and IGF. The mention of ‘adapting WSIS action lines’ has been deleted from para. 69, as have the specific mentions of ‘internet governance’ in para. 73 on the IGF. Instead, there is a suggestion for IGF to add an ‘annual policy discussion track’ on GDC implementation.

Why? The simplification of follow-up procedures seems to respond to calls from Member States during the consultations to make maximal use of existing structures. Rev.1 tries to clarify what the vision is for WSIS and IGF and how important the two processes should be in the GDC’s implementation and follow-up. The change with regard to WSIS reflects, perhaps, an unwillingness to renegotiate the core WSIS texts through the WSIS+20 review, and rather use what is already there in its current form. On the other hand, where IGF is concerned, deleting the references to ‘internet governance’ seems to broaden the envisaged role of IGF, implying that it is relevant to all digital issues and should take a full and active role in GDC implementation as a kind of multistakeholder oversight body. 

Other things to note:

  • The proposal for a Global Fund for AI and Emerging Technology remains almost identical to the zero draft. 
  • The proposal for an Office within the UN Secretariat for digital and emerging technologies is also not substantially changed, although Rev.1 makes it explicit that this would be an institutionalization of OSET, and also deletes the mention of a presence of this office in Geneva.

In conclusion, at least as far as the governance of AI and emerging technology is concerned, this Rev.1 seems to be a step in the right direction as it simplifies and re-structures, without substantially reducing the ambition of the Zero Draft. This will hopefully provide more time for Member States to process the institutional proposals and get ready for the line-by-line negotiations in a few weeks.