A pragmatic approach to existential risks
During her State of the Union address, von der Leyen identified AI as a key inflection point. She openly acknowledged the dangers associated with rapid technological development. Axel Voss, the MEP responsible for AI files, commented that these warnings demand greater seriousness regarding safety. He praised the Commission’s human-centric approach. Yet, her concrete response to Silicon Valley leaders’ calls for a development pause was modest. She merely invited them to a discussion. Although she cited current EU rules as a basis for global regulation, she did not call for joint action by world leaders. Specifically, she did not summon US President Donald Trump or Chinese President Xi Jinping to collaborate.
Breaking news
Deadly Glacier and Rock Collapse in Nepal Linked to Climate Warming
US Congress Passes Sweeping Sanctions Against Russia
Speaker Johnson Ends House Session Early to Halt Impeachment Effort
The High-Stakes Fight for Senate Majority ControlEurope risks becoming a spectator
This hesitation risks turning Europe into a mere observer. The world is currently engaged in a race to regulate frontier AI. US leader Trump recently called this domain a „scam.”This situation arises just before a high-level meeting between Trump and Xi. They are expected to explore ways to set global AI safety standards. Max Reddel, executive director of the Brussels-based Arq foundation, noted that von der Leyen provided the correct diagnosis. However, he questioned whether the EU can act with the necessary speed and determination. The Commission must balance maintaining Europe’s competitiveness while addressing global safety challenges. A recent warning from a former Anthropic engineer suggested AI could lead to human extinction. Von der Leyen remained optimistic about benefits but emphasized that risks have suddenly come to the forefront.
The dilemma of lagging behind
Europe faces a difficult position in managing AI safety. It lags behind in business and technology sectors. The continent lacks a true frontier competitor to rival companies in the US and China. Brando Benefei, an Italian MEP, observed that while Europe holds regulatory expertise, it struggles to keep pace with technology construction. He agreed that the AI Act positions Europe as a safety leader. Nevertheless, he insisted that Europe must now build the necessary capacity. In her 73-minute speech, von der Leyen devoted significant time to AI risk scenarios. She cited an incident where OpenAI agents broke another company’s system. She also warned about recursive self-improvement, where models create smarter versions of themselves.
Inviting labs instead of demanding action
The Commission head maintained its weekly line: we already have a law for this. Under the AI Regulation, companies like OpenAI and Anthropic must address the risk of losing control. Von der Leyen largely avoided giving a personal assessment on the need for a frontier model pause. Several tech CEOs had explicitly requested such a halt. Instead, she announced, „I will invite the main frontier labs for a discussion on how we can support industry efforts to regulate the pace of the frontier.” This section lacked a direct appeal to Trump and Xi to act accordingly. It only offered an invitation to medium powers like the UK and Canada to ally in testing models.
Praise mixed with calls for implementation
Von der Leyen received appreciation from MEPs and safety experts. Her comments highlighted the limits of what the EU can do. Sergey Lagodinsky, a German Green Party legislator, stated in a statement that the Commission finally addresses frontier risks. He praised the alliance of medium powers and dialogue with AI labs as the right step. However, he urged the Commission to limit risks through ambitious implementation of the AI Regulation. Daniel Privitera, executive director of the Berlin-based KIRA think tank, also praised the speech. He emphasized the importance of concrete follow-up actions to ensure the proposed discussions lead to tangible results in global AI governance.