New Delhi: US-based artificial intelligence (AI) company Anthropic has withheld its latest AI model from pre-release safety testing by the United Kingdom’s AI Security Institute (AISI), says a Financial Times report. This is reportedly the first time that AISI has been kept out from an Anthropic pre-release testing exercise.
Anthropic is a San Francisco-based AI company behind the Claude family of AI models. On 1 September, the company launched Mythos 5.1 and Fable 5.1, the latter available to the general public, and the former a restricted version strictly designed for select cybersecurity and life-sciences partners.According to the FT report, authorised US organisations received access to Mythos 5.1, but the UK institute did not. Anthropic has not explained its decision.
The move has raised major concerns within the UK government about cooperation between major AI companies and international safety organisations. The Financial Times reported that officials fear that technology companies could be moving towards a more protectionist approach to advanced AI development.AISI also has previously worked with AI companies to test powerful models before their release to the public. Such evaluations are aimed at identifying risks, including cybersecurity threats and potential misuse of advanced AI systems.Anthropic has generally been cautious towards the approach of releasing its most advanced models. However, this decision comes amid intense competition with companies such as Google and Meta.The move could affect Anthropic’s position in the global AI market race. Markets and researchers are now keenly awaiting further announcements on its models: their performance, availability, safety testing and future ties with international AI security bodies.Developments in the company’s future ties with international AI safety bodies will also be influenced by how governments and the industry view Anthropic’s approach to advanced AI.(Edited by Nardeep Singh Dahiya)
