Warning Shot: Making Sense of 2026 AI Incidents

Are AI companies losing control of their agents? In just the past few months, 700+ OpenAI agents escaped their testing environment and coordinated an attack on another AI company, stealing the answers to their test and attempting to cover their tracks. Other agents accessed confidential information from the Australian medical system, and OpenAI agents even hacked into their own company's internal infrastructure. Join us for an evening of talks, Q&A, food, and networking to explore what happened, the safety implications for the public, and what can be done to mitigate AI risks.Registration is free. No technical background required.— Programme Talk 1 | Stanislav Lukyanenko: AI engineer, independent researcher Rebuilding the timeline of the “Hugging Face incident:” how did the self-titled agent swarm manage to coordinate, escape their environment, and hack into another AI company? Stanislav will reconstruct the extraordinary sequence of events and discuss what the evidence says about AI behaviour. Talk 2 | Corvin Binder: AI governance expert, CEO Conformance AI What do these incidents mean for the future of AI safety? Corvin will discuss why this matters for businesses, governments and the public, and where this is heading from an AI governance perspective. — After the talks, join a Q&A with both speakers, followed by informal networking with catered food and soft drinks. Good to know No technical background required. Everyone curious about AI and its impact on society is welcome. Free to attend, but registration is required and seats are limited. The event will be held in English, and the talks will be recorded. —Disclaimer: the views presented in this event do not represent the views of Netlight. Hosted by Rationality Munich.