EU AI Safety v0.5 Informational Event

Evaluating whether AI systems are safe is a crucial challenge for policy makers, technology companies, and society as a whole. The AI Safety Working Group [ one pager, website ] is working to create an industry-standard safety testing platform to realize the benefits while minimizing the risks.

We would like to invite you to an informational event organized by the MLCommons Association [ one pager, website ] in support of reaching our v0.5 platform milestone. The event will be held at the Sparks event space in Brussels on April 18th and 19th, with a full day’s agenda on the 18th, followed by a social dinner on the evening of the 18th, and a half-day morning event on the 19th. 

Our goal is to bring together global thought leaders to collaborate with us on this important task. We believe that AI, similar to other advanced technologies such as cars, airplanes, and medicines, needs standard safety benchmarks to achieve three critical goals:

1.      Guide responsible development and facilitate research

2.      Support consumer/purchaser decision making

3.      Enable technically sound and risk-based policy regulation

We previously hosted two AIS kickoffs in the US and EMEA, with over 100 attendees and 50 organizations to both events combined. We’re keen to share our progress to date with an informational event to share progress towards our v0.5 platform milestone and seek input from a diverse set of stakeholders into our next stage of development. 

We especially welcome people involved in policy and standards discussions globally who can help shape the benchmarks into a rigorous yet practical toolbox for supporting their goals, as well as subject matter experts located in EMEA who may want to contribute to benchmark, test, platform, or engine development, as well as provide use-case input and design.

The Brussels informational event agenda will include:

·       Panels of ecosystem stakeholders weighing in on the utility of benchmarks: policy-makers from different jurisdictions, standards body representatives, AI safety providers, developers, and integrators.

·       In-depth technical review of the safety benchmark and the testing platform roadmap.

·       v1.0 challenges and aspirations and requests for input: benchmark, test, platform, and engine design, as well as community engagement and overall effort governance.


Please RSVP by April 5th. Looking forward to a great discussion!

AI Safety Working Group
Joaquin Vanschoren, Peter Mattson, Percy Liang