News

Warsaw AI Safety Day will take place on October 24, 2026 — the first event in Poland devoted entirely to the safe development of advanced artificial intelligence (AI Safety & Governance). We are planning an event for nearly 250 people. The program focuses primarily on networking, exchanging experiences, building collaborations, mentoring sessions, and discussions with experts. The event is organised by the AI Safety Poland Foundation and SKN EA UW, in collaboration with ML in PL.

Places are limited and every application is subject to review. Lower-priced Early Bird tickets are available until September 20. Applications and further details are available at: warsawaisafety.day

We have just wrapped up a successful meetup organised together with the Centre for Credible AI at the Faculty of Mathematics and Information Science, Warsaw University of Technology. Julia Bazińska spoke about the differences between “AI Safety” and “AI Security” and about vulnerabilities in LLMs, while Jan Piotrowski presented the evaluation of multi-agent systems using the LEAN language.

The second part featured a panel discussion with Julia, Tomasz Steifer and Kacper Dudzic. The experts discussed the widely covered OpenAI x HuggingFace attack, the question of “independent” goals of AI agents, and the growing impact of open weight models on cybersecurity.

Thank you to our speakers, panellists and participants for a great time and fascinating discussions!

Three people from our community have been accepted into the autumn 2026 cohort of the Talos Fellowship: Zuza Topolska, Michał Tarnowski and Paweł Kamiński.

The Talos Fellowship is one of the most important European programs preparing participants for a career in AI policy — it includes an eight-week course on the fundamentals of European AI governance, a week-long summit in Brussels, and internships and grants for participants' own initiatives.

Out of 21 participants from across Europe, as many as three represent the Polish ecosystem, which speaks well of the growing strength of the AI safety and AI governance community in Poland. Congratulations, and we're keeping our fingers crossed for the months ahead!

We are delighted to announce that Patryk Perduta from our community took second place in the Technical AI Safety Puzzle competition organised by BlueDot Impact.

The challenge focused on neural network interpretability. Participants had to discover which feature of text the model was encoding in an unusual, non-linear way, explain the geometry of that representation, and then train their own model using an even more unusual method of encoding information.

Congratulations to Patryk on this excellent result. We encourage everyone interested in this topic to browse the competition tasks — it is a great opportunity to put your skills to the test in practice. Patryk's solution is available here: blog.perduta.net

On 29 June 2026 the next AI Safety Poland Meetup took place at HEVRE in Kraków! Attendance reached 90 people, making this our largest meetup in Kraków to date.

The main highlight of the programme was 5 engaging Lightning Talks with Q&A sessions. The speakers were: Piotr Kędziora, Jakub Nowak, Kacper Dudzic, Alicja Grochocka-Dorocińska and Dawid Rymarczyk.

After the official part came time for networking in a fantastic atmosphere. The event was a great opportunity to integrate our community with participants of the MLSS 2026 summer school. Huge thanks to the ML in PL team for another great collaboration, and to everyone who attended — thank you for the incredible energy!

We are proud to wrap up a series of 17 biweekly webinars from the AI Safety Poland Talks cycle. Over the past months we have created a space for expert discussions on the safety, ethics and risks associated with the development of artificial intelligence. We thank all the speakers and participants for their engagement and for building the community together.

During our last meeting we hosted Krzysztof Bar, CEO of the London Initiative for Safe AI (LISA), who shared his career path from quantum computing to his involvement in AI Safety. During the Q&A session we discussed the work of this key European hub and analysed the biggest challenges in AI Safety.

We are delighted to announce that on 11 June 2026 the AI Safety Poland community was formally registered in the National Court Register (KRS: 0001247485) as a foundation, with its registered office at ul. Św. Marcin 29/8 in Poznań.

Strategic oversight of our activities will be handled by the newly appointed Supervisory Board, composed of: Chris Szulc, Jakub Kryś, Marcel Windys, Patryk Wielopolski and Piotr Kędziora.

This historic step for our community opens entirely new opportunities for the Management Board, led by Jakub Nowak, to professionalise our work on the safety of artificial intelligence in Poland.

Piotr Kędziora, together with two members of his team — Miłosz Bruski and Rwiti Bhattacharya — won the BlueDot AI Risk Content hackathon yesterday, having flown to London from Kraków especially for the occasion.

In its verdict, the competition jury particularly recognised his exceptionally creative approach to the topics tackled. We encourage you to watch all the awarded works.