Twelve days after the EU AI Act’s enforcement machinery became active, the world’s oldest AI research conference opens in Bremen — and a European Commission technology specialist will explain how academics and industry can shape the law’s implementation. The Article 50 transparency obligations of the EU AI Act became enforceable on 2 August 2026, and Jeroen Delfos of the European AI Office is scheduled to speak at IJCAI‑ECAI 2026 on Wednesday 19 August at 11:30 a.m. CEST in the Hanse room at the Congress Centrum Bremen.
EU AI Act enforcement arrives as IJCAI convenes
With Article 50 in force, any operator running a chatbot for EU audiences must disclose at the start of every interaction that users are communicating with an AI system. Providers of generative systems are required to embed machine‑readable provenance markers in synthetic audio, images, video and text; deepfakes must be labelled. Non‑compliance carries fines of up to €15 million or 3% of global annual turnover, whichever is greater. The European AI Office gained full enforcement powers over general‑purpose model providers, including the ability to levy fines.
Delfos, a technology specialist involved in developing standards, Commission guidelines and coordinating market surveillance for enforcement, will present a talk titled The EU AI Act — Help Shape the Future of Safe AI. His session runs in parallel with the main technical program, offering an opportunity for researchers to hear from an enforcement official about ongoing standard‑setting and market surveillance work.
Conference scope, schedule and flagship sessions
IJCAI‑ECAI 2026 opens its workshop and tutorial program on Saturday 15 August at the University of Bremen campus; the main technical conference runs Tuesday–Friday 18–21 August at the Congress Centrum Bremen. Organizers expect around 4,000 researchers, engineers and practitioners from more than 40 countries. IJCAI, founded in 1969, is the oldest continuously running AI research conference and this edition accepted 990 papers across all tracks, including 713 in the Main Track.
The invited keynotes span multiple AI traditions. Tuesday opens with Nick Jennings of Loughborough University on The Past, Present and Future of Agentic AI at 10:00 CEST, followed by Luciano Serafini on neuro‑symbolic approaches and Logic Tensor Networks. Wednesday begins with Barbara Plank at 9:00 CEST on The Myth of Ground Truth; Jeroen Delfos speaks at 11:30 a.m. CEST; and David C. Parkes accepts the John McCarthy Award with a keynote from 14:00–15:00 CEST. Thursday features Tinne Tuytelaars and the Computers and Thought Award presentation to Jiajun Wu; Friday closes with Jan Peters and Doina Precup.
A three‑day pre‑conference period (15–17 August) hosts 40 workshops and more than 25 tutorials. Notable workshops include the Safe Agentic AI Framework and Ecosystem Roadmapping (SAFER) on 16 August, RobustifAI on 15 August, and the inaugural IJCAI Workshop on Safe Physical AI (15–16 August). Transparency and explainability feature in two separate two‑day workshops: the IJCAI 2026 Workshop on Explainable Artificial Intelligence (XAI) and the CLEAR Workshop. Policy‑oriented sessions include Augmented Democracies: AI for Civic Engagement and Democratic Decision‑Making (17 August), Value Engineering in AI (VALE) on 15 August and a Workshop on the EU AI Act on 19 August. A public outreach panel, New Directions for Safe AI Agents, is free to attend with registration on Sunday 16 August at 18:00 CEST at the University of Bremen.
Neuro‑symbolic topics are well represented in workshops such as Neuro‑Symbolic Intelligence for LLMs and Autonomous Agents (NILA) on 16 August and Logical and Symbolic Reasoning of Large Language Models on 17 August.
International participation, governance context and Bremen’s role
IJCAI’s non‑profit structure and governance under German law are highlighted this year amid a wider fracture in the global AI conference circuit. In March 2026, NeurIPS included a US Treasury sanctions compliance clause in its submission handbook, prompting a boycott call from the China Computer Federation and a decision by the China Association for Science and Technology to stop funding Chinese scholars’ attendance at NeurIPS and to strip NeurIPS papers of recognition in CAST‑affiliated evaluations. NeurIPS later revised the clause and issued an apology, but CAST’s measures remained in place.
IJCAI is organized outside US legal jurisdiction and this edition’s sponsor list confirms active Chinese institutional participation: the School of Artificial Intelligence at the Chinese University of Hong Kong (Shenzhen) is listed among academic supporters, tutorials feature researchers from Tsinghua University and Zhejiang University, and Alibaba researcher Jingren Zhou is on the program. The conference therefore provides a venue where researchers from a range of institutions can appear in the same proceedings.
Bremen was chosen in part for its local AI infrastructure. The German Research Center for Artificial Intelligence (DFKI) operates a major site in Bremen, the University of Bremen coordinates the U Bremen Research Alliance linking thirteen research institutes, and the EASE collaborative research center supplies a hands‑on tutorial in virtual cognitive robotics. The IJCAI Industry Day on Wednesday 19 August at Messe Bremen includes presentations from DeepL, T‑Systems and SAP, plus an interactive startup session featuring three Bremen startups: Count from Above, ellamind and prokube.
IJCAI‑ECAI 2026 runs 15–21 August 2026 at the University of Bremen and Congress Centrum Bremen. The full program is available on the conference website; Delfos’s EU AI Act session is in the Hanse room at 11:30 a.m. CEST on 19 August, and the public outreach panel on Safe AI Agents takes place at 18:00 CEST on 16 August at the University of Bremen.
