New International Agreement Targets Artificial Intelligence Safety

At the initiative of French President Emmanuel Macron, a new international agreement on artificial intelligence safety has been signed in Paris. Known as the "Paris Call for AI Safety and Security," the document was signed by ten of the world’s leading technology companies, including OpenAI, Anthropic, Microsoft, and Mistral. The meeting, held as part of the Paris Peace Forum, brought together state leaders, industry experts, and tech executives, including Sam Altman and Dario Amodei. ## Commitments Under the Agreement Under the terms of the agreement, signatories have committed to developing "fail-safe" mechanisms that would allow for the immediate shutdown of AI models if necessary. This applies to both the training of models and their practical deployment. Furthermore, companies have pledged to provide greater transparency regarding system capabilities and to allow independent, qualified experts to audit their technologies. The declaration also includes a shared approach to risk management, aimed at preventing threats such as cyberattacks, the development of weapons, or the loss of human control over systems. Crucially, the parties agreed to share early warnings in the event of serious incidents. ## Criticism and Shortcomings Despite this significant step, the agreement has faced criticism. The French AI safety organization CeSIA, a signatory to the document, stated that the agreement "lacks teeth." Critics argue that the declaration fails to establish clear thresholds for risk assessment to determine when a model is safe for release. There are also no guarantees that independent auditors will have full access to the systems. The organization maintains that the state should hold real authority to halt a model if its risks exceed acceptable norms. Notably, tech giants such as Google and Meta did not join the agreement, and the United States, despite active participation in discussions, did not sign the declaration. ## Focus on Child Safety During the forum, Emmanuel Macron placed particular emphasis on the impact of AI on children. He urged the industry to subject any AI tool accessible to minors to rigorous pre-certification. The President compared this process to safety testing for toys, noting that the influence of chatbots on children's mental health has already become a serious challenge. While the Paris Declaration remains open to new members, it lacks specific timelines and a concrete action plan. For users in Georgia, these developments are significant, as the establishment of international standards will directly define the safety frameworks that will govern digital services available in the country in the future.