Anthropic's AI Model 'Mythos' Ignites Controversy Over Transparency and Investment Motives
This week, the AI company Anthropic announced it had developed an AI model so powerful that, citing overwhelming responsibility, it would not release it to the public. The model, named Mythos, was described as posing catastrophic cybersecurity risks, prompting the US treasury secretary, Scott Bessent, to summon major bank heads for discussions, and Reform UK MP Danny Kruger to urge the UK government to engage with Anthropic.
Scepticism and Accusations of Hype in the AI Industry
However, the announcement has been met with significant scepticism from experts and critics. Noted AI critic Gary Marcus commented, "Dario [Amodei] has far more technical chops than Sam [Altman], but seems to have graduated from the same school of hype and exaggeration," referring to the CEOs of Anthropic and its rival, OpenAI. This sentiment echoes broader concerns about whether Anthropic's move is a genuine safety measure or a strategic marketing effort to garner investment.
Dr Heidy Khlaaf, chief AI scientist at the AI Now Institute, highlighted the lack of substantiated evidence, stating, "Releasing a marketing post with purposely vague language that obscures evidence ... brings into question if they are trying to garner further investment without scrutiny." This raises doubts about the transparency and intentions behind Anthropic's claims.
Anthropic's Media Strategy and Public Perception
Amidst this controversy, Anthropic has been actively cultivating a media presence. In recent months, the company has been featured in extensive profiles in the New Yorker, Wall Street Journal, and on the cover of Time magazine, with CEO Dario Amodei depicted in a dramatic style. Anthropic's media lead, Danielle Ghiglieri, has celebrated these achievements on LinkedIn, emphasizing the company's narrative control.
Other tech public relations professionals have taken note, with one remarking, "They are clearly having a moment right now but companies building technology that will change the world deserve equal scrutiny." This comment followed an incident where Anthropic accidentally leaked part of Claude's internal source code in early April, though the company assured no sensitive data was exposed.
Technical and Resource Constraints Behind the Decision
Beyond the hype, practical reasons may have influenced Anthropic's decision not to release Mythos. The company is reportedly struggling with limited computing resources, having introduced usage caps on its popular Claude model and requiring extra payments for third-party tools. Jameison O'Reilly, an expert in offensive cybersecurity, acknowledged Mythos as a real development but questioned the significance of some claims, such as finding thousands of zero-day vulnerabilities, noting their limited impact in real-world scenarios.
O'Reilly explained, "In those 10 years, across hundreds of engagements, the number of times we needed a zero-day vulnerability to achieve our objective was vanishingly small." This suggests that Anthropic's cybersecurity justifications might be overstated.
The Broader AI Race and Ethical Implications
Anthropic, like OpenAI, is in a fierce race to raise billions and capture a market for AI assistants in personal and corporate settings. Differences between AI products often hinge on intangible attributes like "sense of self" or "soul," making the battle for public trust and investment crucial. Khlaaf warned that this could mirror OpenAI's approach, where safety is used as a PR tool before prioritizing profits, with Anthropic potentially obscuring this shift more effectively.
As protesters in San Francisco recently called for a pause in AI development, marching to Anthropic and OpenAI offices, the debate underscores the tension between innovation, safety, and commercial interests in the rapidly evolving AI landscape.
