US signs AI security order as EU staffs enforcement bodies
A White House executive order creates a voluntary frontier-AI framework; the EU appoints 234 experts to enforce the AI Act.
By BINA Editorial
Today both sides of the Atlantic moved to put institutional weight behind AI policy.
White House signs executive order on AI innovation and security
On 2 June 2026, President Trump signed an executive order titled "Promoting Advanced Artificial Intelligence Innovation and Security." The order establishes a voluntary framework under which developers of frontier AI models are invited to submit systems to the federal government 30 days before public release for a security review. Federal agencies are directed to prioritise cyber defence of national security systems, and the Attorney General is asked to pursue criminal use of AI for hacking and data theft. The order explicitly states it does not authorise mandatory licensing or permitting — a provision designed to assure the industry that innovation will not be bottlenecked by regulatory approval.
EU Commission appoints 234 independent experts to enforce the AI Act
The European Commission announced on 1 June 2026 the formal appointment of two bodies to support enforcement of the EU AI Act. A 60-member Scientific Panel will focus on general-purpose AI models, assessing systemic risks and supporting cross-border market surveillance. A larger Advisory Forum of 174 members — drawn from academia, civil society, industry, SMEs, and startups — will advise the Commission's AI Office and national authorities on standards and implementation questions. Both bodies serve two-year terms and complete the enforcement infrastructure for the GPAI layer of the Act, which has been in force since August 2025.
EU AI Act adds ban on AI systems that generate non-consensual intimate images
The Digital Omnibus agreement, finalised in May 2026, adds a new absolute prohibition to the AI Act: placing on the European market any AI system that generates or manipulates realistic depictions of identifiable people in sexually explicit situations without their explicit consent. A parallel ban covers AI systems that produce child sexual abuse material. Both prohibitions take effect on 2 December 2026 and carry fines of up to €35 million or 7% of global turnover for violations.
AI data centres' power demands strain grids and face political pushback
The rapid expansion of AI infrastructure is meeting a hard constraint: electricity. Reporting from Arizona and across the western United States describes how the next generation of AI data centres requires significantly more power than earlier facilities, threatening to raise household electricity bills and expand the sector's carbon footprint. In Arizona, local officials and community activists are blocking new construction permits despite a prior Nvidia and TSMC commitment of $600 billion to build AI chip manufacturing capacity in the state. Similar friction is emerging in European markets, where governments are balancing AI investment ambitions against renewable energy obligations.
Claude Opus 4.8 becomes the first model to break the 60-point benchmark barrier
Anthropic's Claude Opus 4.8 reached a score of 61.4 on the Artificial Analysis Intelligence Index v4.0 as of 1 June 2026 — the first model in the index's history to exceed 60 points. It also leads the GDPval-AA benchmark for real-world economic tasks with an Elo rating of 1,890, surpassing the next model by 121 points. Gemini 3.1 Pro retains the lead in multimodal capabilities and long-context tasks. The results confirm a widening gap between frontier models and the broader field.