We’re looking forward to the third AI Action Summit in Paris next week and in particular, to discussing how artificial intelligence and the innovations and breakthroughs AI will launch can drive economic growth.
To support and inform this conversation, a particularly important one for Europe, we’re bringing our “AI Economics” event series to Paris, convening policymakers and leading thinkers on the economics of innovation and the importance of making sure AI benefits everyone. After this event, and similar AI Economics gatherings in London and Brussels later this month, we will publish our Economic Blueprint for Europe with recommendations on how to seize the AI opportunity.
And as we reflect on the progress made since Bletchley and Seoul, we are proud to showcase our ongoing commitment to safety, transparency and accountability in shaping the future of frontier models.
Between the hot topics of safety and growth, we are confident that Paris will be another important milestone towards the responsible and beneficial development of AI for everyone.
Unlocking the opportunities of AI
Today, people are using AI tools at home and at work more than ever before, with 300 million weekly active users globally on ChatGPT alone. As societies, we are experiencing first-hand how AI can be used to assist with everyday tasks and its potential to help us solve difficult problems.
In France we are witnessing how people, businesses, and academic institutions are using AI to solve complex problems, boost productivity, foster creativity, and even advance scientific discoveries:
- Sanofi is using our AI to accelerate drug trials, helping life-saving medicines reach patients faster
- the ESCP Business School is offering personalised learning experiences
- Orange will fine-tune our open-source speech model to better understand regional languages in Africa
- Ask Mona is promoting greater access to culture with their museum experiences
Earlier this year OpenAI released a US Economic Blueprint that lays out our policy proposals for capitalizing on the benefits of AI and driving economic growth across communities. This is a program that we are working to adapt and roll out in other countries.
The summit’s wider focus on opportunities and innovation will also shine a spotlight on the startups and enterprises that are adopting AI to boost productivity and create new, innovative solutions. Nowhere is this momentum more evident than at Station F, the world’s largest startup campus, where we will take part in the summit’s official Business Day.
We are proud to partner with innovative French startups like Mirakl and Pigment, who are using our AI to compete and grow on a global scale, and look forward to engaging with other businesses that are driving AI adoption in Europe. For OpenAI, this is an opportunity to deepen our ties with the ecosystem in France following the opening of our first continental European office in Paris last year.
Update on voluntary commitments
OpenAI is a signatory to both the Bletchley Declaration(opens in a new window) (November 2023) and Seoul Framework(opens in a new window) (May 2024) which focus on mitigating risks associated with foundation models, accountability for their safe development and transparency on safety measures.
Following the Bletchley summit we published our Preparedness Framework(opens in a new window) which outlines our approach to deploying frontier AI models safely. The framework is grounded in science-driven measures, iterative deployment, and continuous tracking across four key areas, aligning with our commitments from Bletchley and Seoul: cybersecurity, chemical, biological, radiological, and nuclear threats (CBRN), persuasion, and autonomy, as well as continuous discovery of “unknown unknowns,” with risk thresholds ranging from low to critical.
In keeping with our commitment to the Seoul Framework and its requirement to continuously review internal accountability and governance frameworks, we are currently working towards an updated version of our Preparedness Framework that will be published later this year. This update will reflect refinements to our risk thresholds, mitigation strategies, and more. We look forward to incorporating insights from the Paris summit as we finalize this work.
In addition to the Preparedness Framework we continue to advance our safety initiatives through a range of rigorous internal processes, external collaborations and transparent reporting, including:
- Publishing detailed reports, called system cards, for new frontier AI systems that we deploy. Since Seoul, this includes system cards for five of our frontier models: 4o, o1, Sora, Operator, and o3-mini. We will share our safety insights and safeguards for deep research in a system card before we expand access.
- Establishing an internal governance structure to uphold procedural commitments and ensure effective risk management.
- Developing and engaging with an extensive Red Teaming Network which consists of individual experts, research institutions, and civil society organizations in a wide variety of domains who help identify risks across CBRN threats and societal harms.
- Agreements with the US & UK AI Safety Institutes that have since conducted capability evaluations on our o1-preview and o1 models.
In anticipation of the summit, we have prepared a more detailed report showing the full extent of our work on the voluntary commitments and safety here(opens in a new window).
We look forward to sharing our progress on safety and engaging with world leaders and civil society groups in Paris, and are committed to continuing to fulfill the Bletchley and Seoul voluntary commitments in future.
