The US government is urging Meta to submit its models for review due to concerns about security risks posed by artificial intelligence technologies. According to The New York Times, Meta is the only major artificial intelligence developer that does not willingly open its models to review.

In particular, government officials aim to evaluate the capabilities of Meta’s artificial intelligence models and identify possible vulnerabilities. Other major players such as OpenAI and Anthropic continue to collaborate with the government to test their now unreleased models.

Government’s AI review process

Companies such as Google, xAI and Microsoft have reached an agreement with the Artificial Intelligence Standards and Innovation Center to provide early access to their new models. Led by Commerce Secretary Howard Lutnick, this agency operates a team of technical experts to oversee artificial intelligence technologies.

Meta spokesman Francis Brennan stated that the company shares the goals of developing safe artificial intelligence and is working on agreement details. It is said that the government conveyed these demands to the company via e-mail and the process is expected to be completed in a short time.

An executive order signed by President Trump on June 2 authorized the federal government to create a framework for evaluating versions of artificial intelligence. The government is scheduled to develop a review process by the end of July, and the aim is to give authorities a 30-day evaluation period before companies offer their technologies to the public.

Increased security audits in the industry

Meta released the Muse Spark model with “Instant” and “Thinking” modes in April. Thinking mode allows the model to make judgments before responding so it can provide more comprehensive answers.

The government closely monitors the artificial intelligence sector and increases security measures. In mid-June, Anthropic was ordered to suspend foreign-born individuals’ access to the Mythos 5 and Fable 5 models, citing national security concerns.

In order to comply with this directive, Anthropic has completely closed access to the models in question. Mythos 5 stands out as a cybersecurity-focused model that the company offers exclusively to Project Glasswing partners. Fable 5 stands out as a version developed to bring the capabilities of this model to a wider audience.

Do you think governments controlling artificial intelligence models in this way is a necessary step for the technology world, or will it slow down innovation?