Precedent
The NPT has a rigid amendment process, requiring approval by “a majority of the votes of all the Parties to the Treaty.” This intentionally makes formal changes extremely difficult. Our treaty follows this precedent with the aim of fortifying the agreement against short-term pressures to relax thresholds or weaken provisions.
Hard-to-amend (and thus hard-to-weaken) treaties rely on other mechanisms for strengthening as needed. The NPT has never been amended, but has been adapted through the five-yearly Review Conference stipulated in Article VIII, where consensus agreements are made “with a view to assuring that the purposes of the Preamble and the provisions of the Treaty are being realised.”
Similarly, Article XII of the 1975 Biological Weapons Convention relies on its five-yearly Review Conferences to strengthen the treaty through non-binding Confidence-Building Measures, as formal amendments are rare. Our agreement stipulates a three-year conference, as AI has been a field prone to rapid shifts; this period may need to be further shortened.
Article XV of the Chemical Weapons Convention makes a distinction between amendments and administrative or technical changes, with less stringent approval provisions for the latter. Similar language could be added to our draft agreement to provide a level of flexibility in managing future developments in the field of AI.
Article XV of the Outer Space Treaty Article contains an amendment clause, but the treaty has never been formally amended; instead, new treaties have been negotiated to address emerging space issues. This could be another option for shoring up weaknesses that may become apparent in an AI treaty.