We’re running out of terminology! In the last couple of weeks the EU Commission initiated its AI Pact to get industry to implement the AI Act which is not yet finalized and a group of 18 countries signed a pact on AI aimed at keeping artificial intelligence safe from rogue actors.
Here’s how they differ:
Pact to make AI ‘secure by design’
On Sunday 18 countries, including the U.S., Germany and the U.K., signed a pact aimed at making AI secure by design. The Guardian reports that the group of nations (including Germany, Italy, the Czech Republic, Estonia, Poland, Australia, Chile, Israel, Nigeria and Singapore) signed the pact which involves measures like maintaining the security of data and making sure that trustworthy parties are involved in supplying the technology behind AI functionality.
The agreement lacks binding force and primarily includes broad suggestions, such as overseeing AI systems to prevent misuse, safeguarding data against manipulation, and thoroughly evaluating software providers. It also deals with questions of how to keep AI technology from being hijacked by hackers and includes recommendations such as only releasing models after appropriate security testing.
European Commission AI Pact
On 15 November, the EU Commission initiated the AI PACT which it had been working on since the spring. The AI Pact seeks voluntary commitments industry to anticipate the AI Act and to start implementing its requirements ahead of the legal deadline. Whereas certain aspects of the AI Act will take effect shortly after the Regulation is adopted, others, such as specific requirements about high-risk AI systems, will only become applicable after a transitional period. The Commission does not want to wait that long and is calling on industry to start applying the principles contained in the AI Act (which is still being negotiated…). Back in May, Thierry Breton (European Commissioner for Internal Market) said “Sundar ( CEO of Google) and I agreed that we cannot afford to wait until AI regulation actually becomes applicable, and to work together with all AI developers to already develop an AI pact on a voluntary basis ahead of the legal deadline".
The Pact consists of industry commitments which will take the form of pledges to work towards compliance with the upcoming AI Act accompanied by details about concrete actions being carried out or planned to address specific requirements of the future AI Act.
The Commission is selling the Pact as a way to allow front-runners to test and share their solutions with the wider community, giving them a first-mover advantage. The Commission will work together with the participants and support them in:
building of common understanding of the objectives of the future AI Act;
taking concrete actions to understand, adapt and prepare for the future implementation of the AI Act (e.g. build internal processes, prepare staff and self-assess AI systems);
sharing knowledge and increase the visibility and credibility of the safeguards put in place to demonstrate trustworthy AI; and ultimately
building additional trust in AI technologies.
Industry players interested in the scheme can now express their interest in the scheme. As a next step, the Commission will bring together interested parties, in the first half of 2024, to discuss the ambitions of the Pact and collect preliminary ideas and best practices which could inspire future pledges.