Antonio Tejeda Encinas | President of the Euro-American Committee of Digital Law - CEA Digital Law | CEO of META Channel Corporation
“SB-1047 veto sparks clash of giants in the global AI regulatory race”
The Governor of California’s veto of bill SB-1047, the “Safe and Secure Innovation for Frontier Artificial Intelligence Models Act”, marks a key moment in the debate on regulating artificial intelligence (AI) in the United States. Given the growing global interest in regulating advanced AI technologies, it is worth comparing the framework proposed in California with the European Union’s Artificial Intelligence Act (AI Act) and the Council of Europe Convention on Artificial Intelligence, Human Rights, Democracy and the Rule of Law.
California bill SB-1047: a narrow focus and business opposition
SB-1047 was designed to impose significant obligations on developers of so-called frontier models, defined as those requiring large amounts of computing power. These obligations included:
1. Periodic safety assessments:
Implementation of “emergency shutdown” mechanisms, or “kill switches”.
2. External audit requirements:
Restrictions to prevent use in high-risk settings without prior assessment.
Despite the backing of companies such as Anthropic, and of Elon Musk, it was vetoed over concerns that it would slow innovation and apply overly stringent standards even to low-risk uses.
The European Union’s AI Act: a risk-based approach
The EU Artificial Intelligence Act (AI Act), proposed in 2021 and adopted in 2024, with gradual application from 2025, establishes a comprehensive risk-classification system for AI technologies:
High-risk AI: includes applications that affect people’s safety or fundamental rights, such as credit-scoring systems, surveillance algorithms and medical AI applications.
Outright prohibitions:
certain AI systems are banned, such as those that enable mass surveillance or manipulate people’s behaviour against their will.
Additional requirements for general-purpose AI and widely used models, such as transparency on training data, mandatory audits and cybersecurity requirements.
The EU Regulation is much broader and classifies by level of risk rather than focusing on the size or computing power of models, as SB-1047 did. The Regulation also calls for assessment based on the specific use of the system rather than the general capabilities of the model, which addresses some of the criticisms Newsom raised in his veto.
Council of Europe Convention: protecting human rights and democracy
The Council of Europe Convention on AI, Human Rights, Democracy and the Rule of Law, adopted in May 2024, focuses on ensuring that AI is developed and used in line with the fundamental principles of respect for human rights and the rule of law. Its distinctive features include:
- The requirement that all AI systems respect human rights by design.
- Mandatory impact assessments before deploying AI in any critical environment.
- A strong focus on transparency and accountability:
Although the Council of Europe Convention is broader in scope, it shares with the EU AI Act a priority on protecting fundamental rights and minimising potential harm to democracy and the rule of law. Unlike SB-1047, it focuses not on the computing power or size of models but on how they may affect society and how they should be designed to align with ethical and legal principles.
Implications of the veto and future trends
The SB-1047 veto reflects a cautious stance by the State of California, keen not to limit the competitiveness of local technology companies. However, the lack of a coherent federal regulatory framework could allow other jurisdictions, such as the EU, to take the lead in AI regulation and impose their rules globally through their extraterritorial reach, as is the case with the General Data Protection Regulation (GDPR).
The debate over AI regulation in California also highlights how hard it is to balance innovation with safety and fundamental rights. While the EU has taken a broad approach based on risk assessment and proportionate regulation, SB-1047 sought direct, preventive safeguards for the most powerful models, which is seen as a more restrictive measure that may not be the best fit for an environment of rapid technological development.
The SB-1047 veto leaves California with a regulatory gap, while the EU and the Council of Europe move towards a more structured and detailed regulatory environment. Whether the US regulates in future will depend largely on collaboration between the private sector, legislators and federal regulators to avoid a fragmented landscape that hampers innovation without ensuring AI safety.
النشرة
فعاليات وأخبار وتحليلات في القانون الرقمي تصلك على بريدك، ولا نراسلك إلا حين يكون لدينا جديد.
