AI governance has been propelled into the limelight—with Members of the European Parliament agreeing on compromise amendments to the EU AI Act, testimonies in front of U.S. Congress, and many scientists signing open letters to signal their concerns regarding the trajectory of AI developments. The pace of developments has kept us busy, and we’re excited to share what we’ve been up to in this quarter’s newsletter:

• Our 2022 Annual Report
• Our EU AI Act achievements
• Takeaways from the fourth edition of The Athens Roundtable on AI and the Rule of Law
• Rwanda approved its National Artificial Intelligence Policy, which TFS supported in drafting
• We’re hiring for a Senior Associate role focused on AI governance in the European Union, based in Brussels. (Applications close June 11th.)

Our 2022 Annual Report

We are pleased to share our 2022 Annual Report. Our 2022 achievements include (but are not limited to!) convening the Fourth Edition of The Athens Roundtable on AI and the Rule of Law at the European Parliament in Brussels; working with legal practitioners and academics from both sides of the Atlantic to publish a Manifesto on Enforcing Law in the Age of AI, calling for the effective and legitimate enforcement of laws concerning AI systems; proposing the design and operationalization of regulatory sandboxes in the EU AI Act; and finalizing national AI strategies and implementation plans for Tunisia and Ghana.

We also take this opportunity to reflect on recent technological developments and chart our course for 2023. We’re continuing work on a number of existing projects, and launching a few new projects that develop, advocate for, and facilitate the implementation of governance mechanisms addressing “general-purpose AI systems” (GPAIS)—large, machine-learning models trained on vast quantities of compute and data that can be adapted to a wide range of tasks and cognitive functions, including some for which they were not intentionally and specifically trained. Due to their versatility, and the advantages they confer in terms of business and productivity, GPAIS have captured not only the attention of technical research communities but have become a new frontier for AI startups racing to develop “artificial general intelligence.”

We believe that the emergence of GPAIS may foreshadow the most disruptive technological revolution in human history. At TFS, we are concerned that these advanced AI systems pose serious risks to human safety and fundamental values. We also believe that relative to the threat that they pose—and in part due to their novelty—GPAIS have been neglected in the AI governance landscape. Thus, in the coming year, we intend to orient our work to focus on the governance of GPAIS, and in particular, towards the research and development phases of such models—stages of the AI lifecycle where risk mitigation is most tractable, and preventative measures would be most effective.

“Will we be able to tame the powerful beasts we are developing and prevent large-scale accidents, misuse, social unrest, or malicious use while allowing innovation to flourish? I am deeply concerned that we still lack adequate institutional instruments, mechanisms, and related capacities and practices. This represents an immense collective action challenge. Given the deep-seated misunderstandings and growing polarization across the AI governance community, forging common ground toward resolute and sustained action will not be easy. It will require humility, dialogue, and willingness to compromise on all sides. We at TFS are committed to it.”

Read more about our 2022 accomplishments and our plans for 2023 in the report accessible via the link below.

EU AI Act Achievements

Earlier this month, Members of the European Parliament approved a draft of Compromise Amendments to the EU AI Act. This latest draft contained a number of measures that TFS has been researching and advocating for, including:

• A special governance regime for foundation models
• The establishment of an AI Office with legal personhood, empowered to update guidelines for foundation models
• National, regional, and EU-wide regulatory sandbox systems
• The establishment of a direct complaint procedure
• The development of tools and infrastructure for benchmarking and assessing dimensions of AI systems
• Whistleblower protections for breaches of Union law by providers of AI systems

With the milestone approval of Parliamentarians and Member States’ versions of the AI Act, we are optimistic that the EU will establish a robust governance framework commensurate with the risks posed by general-purpose AI systems. We are grateful for EU policymakers’ tireless work and look forward to continuing our advisory and collaborations on these issues.

Are you interested in contributing to this work? See our job posting at the bottom of this email!

Rwanda approves National AI Policy

The Cabinet of the Republic of Rwanda has approved the National Artificial Intelligence Policy, which TFS supported in drafting from 2020 to 2021. The Office of the Prime Minister announced the Policy’s approval in a Cabinet resolution communiqué in April 2023.

The drafting of the Policy was an exercise of collective intelligence—led by Rwanda’s Ministry of ICT and Innovation (MINICT) and Rwanda Utilities Regulatory Authority (RURA) and supported by GIZ FAIR Forward, who engaged TFS as implementation partners. As an implementation partner, TFS led a series of workshops with over 120 private sector, public sector, academic, and civil society actors, and validation workshops with the public sector institutions responsible for implementing the Policy. TFS also collected data through interviews with over 35 local experts and 8 surveys administered to workshop participants.

We’re hiring!

We’re expanding our Brussels footprint with a Senior Associate focused on European AI Governance. We’re looking for a highly-motivated individual with experience developing, advocating for, and/or implementing artificial intelligence governance mechanisms in the European Union. This individual will be responsible for leading policy engagement in the EU, including developing, managing, and advocating for recommendations for the EU AI Act, preparing and disseminating policy memos, and building and managing relationships with policymakers, civil society groups, and other relevant actors. Applications close June 11th. We encourage you to share this posting with promising candidates.

Every dollar counts.

The Future Society is an independent nonprofit 501(c)(3) organization. Philanthropy is what makes our work possible. Here’s just one reason to donate to us:

AI governance may be a front-loaded challenge. Laws, policies, and norms have compounding effects. While crafting AI governance will be an ongoing process, getting governance mechanisms in place early is critical. This will necessitate investing heavily in the ‘upfront’ work—conducting research, developing policies and regulations, establishing standards and best practices, and engaging with stakeholders—needed to lay foundations for more robust laws, policies, and norms in the long term, even when many difficult questions remain unanswered. Although this front-loaded effort is onerous, failing to prepare for the technology of the future could prove far more costly.

If that reason wasn’t convincing enough, here are 9 more.