It was a pleasure to welcome a diverse group of more than 100 industry leaders, legal experts and investors to our recent AI Decoded event in the heart of Berlin, where we spent an intensive day exploring how organisations can build, regulate and scale artificial intelligence responsibly – while still keeping up with its blistering pace of innovation.
Trust in AI: Confidence versus responsibility
In our first panel, “Trust in AI – Trust versus responsibility”, Prof. Dr. Tim Stuchtey (BIGS Potsdam / German UDS), Thomas Händl (FTAPI) and Dr. Paul Voigt (Taylor Wessing) examined why trust is becoming a core asset in AI adoption. Prof. Stuchtey highlighted how AI, much like military decision support systems, can compress decision-making time – but stressed that trust lowers transaction costs only where outcomes are reversible and legally manageable. For high‑stakes, irreversible decisions, legal oversight, and clear AI policies are indispensable. Thomas Händl underlined that today’s most valuable assets are digital information, and that most corporate data now flows directly into internet-based AI tools. This creates an inherent tension between trusting AI outputs and maintaining robust controls, especially as no one can fully verify that the AI has been trained on the “right” data. Dr. Voigt warned against over‑reliance on AI output, urging organisations to treat AI as an aid, not an oracle, and to anchor deployment in governance, validation and accountability.
Global legal trends: different routes to regulating AI
The second panel, “International Legal Trends in AI”, drew perspectives from China, Italy, France, the UK and the EU, with contributions from Dr. Jingjing Cao (Taylor Wessing China), Fabrizio Sanna (Orsingher Ortu, Italy), Laura Huck (Taylor Wessing France), Adam Rendle (Winston Taylor, UK) and Thanos Rammos (Taylor Wessing). The discussion made clear that regulatory approaches remain highly fragmented. Dr. Jingjing Cao explained that while China has no overarching AI Act, courts and sectoral rules are already shaping AI practice, with data protection a central pillar and limited scope for opt‑out models. Fabrizio Sanna focused on protection of AI outputs and the growing regulatory and enforcement attention on deepfake tools. From the UK perspective, Adam Rendle described a system without a dedicated AI Act but underpinned by a stringent copyright regime and UK GDPR, where licensing and existing frameworks – including debate around initiatives such as the “Dua Lipa Bill” and automated decision‑making guidance from the ICO – play a decisive role. Laura Huck highlighted that in France, many open questions in AI and copyright now centre on the input side: how training data is sourced, licensed and documented, and how this interacts with the emerging AI Act framework. Across jurisdictions, the panel agreed that companies must navigate converging themes of data protection, copyright and algorithmic accountability, despite very different legislative paths.
Investing in AI: agents, OPCs and the new competitive landscape
Our third panel, “Invest in AI – Use Cases and Investment Trends”, brought together investor and founder insights from Julian Riedlbauer (Drakestar), Thomas Hessler (Ex‑Zanox founder and angel investor) and Dr. Jens Wolf (Taylor Wessing). The conversation underscored that AI-native companies are currently attracting premium valuations, which Julian Riedlbauer argued are justified only where extreme growth rates – sometimes five- to ten‑fold per year – can be sustained. At the same time, many AI business models remain vulnerable, given how quickly they can be copied or displaced by new model releases. He stressed that traditional software businesses must radically transform into AI‑driven organisations to stay competitive.
Thomas Hessler put forward a deliberately provocative thesis: within the next five years, many traditional companies and departments could be overtaken or replaced by OPCs (One Person Companies), in which a single founder orchestrates a fleet of AI agents to perform what previously required entire teams. He framed strategic positioning around an ABCD model: building AI Agents, deploying Bots & autonomous systems, leveraging crypto‑native infrastructure and embracing decentralized energy & compute. Far from being solely a compliance hurdle, he argued that the EU AI Act could become a competitive advantage for transparent, sovereign AI systems that offer full control, auditability and data sovereignty. The panel’s core call to action was clear: organisations should not wait for large‑scale transformation programmes but should identify one recurring process this week and replace it with an AI agent – with clear KPIs, guardrails and human oversight – to start building practical leverage now. The underlying consensus was that the key question is no longer whether companies will use AI, but how quickly and how deliberately they can convert it into sustainable advantage.
Industrialising AI: accelerator or handbrake?
The fourth panel, “Industrialize AI // AI Act – Innovation accelerator or innovation with a hand brake?”, featured Sebastian Gress (Mercedes‑Benz Group AG), Dr. iur. Dipl.-Ing. Lennart Lutz (Cariad SE) and Thomas Kahl (Taylor Wessing). The discussion revolved around the challenge of moving from pilots to production in heavily regulated, safety‑critical industries. While AI promises significant efficiency and performance gains in automotive and mobility, the panel reflected on how the EU AI Act will shape industrialisation – whether as an innovation accelerator that rewards trustworthy, well‑documented systems, or as a perceived brake on speed and experimentation. The conversation highlighted the need to integrate technical engineering, legal risk management and product governance from the outset to ensure that industrial‑scale AI remains both compliant and future‑proof.
Using and developing AI: strategy, employees and roadmap
Our final panel, “Use and Develop AI”, brought together practical perspectives from Timo Wilken (Telefónica), Jan Dröge (DeepL) and Dr. Carolin Monsees (Taylor Wessing). The session explored where AI delivers tangible business value – from efficiency gains and improved quality to risk reduction – and how organisations can translate this into a coherent AI strategy. The panellists compared criteria for in‑house development versus external tools, shared key use cases from legal and operational functions and discussed the concrete steps needed for implementation, from tool selection and risk assessments to documentation of AI use.
A major theme was the role of employees as either drivers or blockers of AI adoption. The panel emphasised the importance of training and upskilling, positioning AI as a way to free capacity for more complex, higher‑value tasks rather than a pure cost‑cutting mechanism. Looking ahead, the speakers reflected on their short‑term plans and five‑year AI roadmaps, including lessons learned on what has worked – and what to avoid – when embedding AI into day‑to‑day business. In light of the AI Act and the Digital Omnibus framework, they also discussed early implementation experiences and how new regulation is shaping product design and compliance-by-design approaches.
Across all panels, one message stood out: AI is moving from experimentation to execution. Organisations that combine trustworthy governance, clear legal frameworks and bold, practical experimentation are best placed to harness its full potential.
