Detailed Analysis
OpenAI CEO Sam Altman publicly criticized Anthropic's marketing strategy surrounding its newly unveiled Claude Mythos AI model, characterizing it as "fear-based marketing" during an appearance on the Core Memory podcast. Altman took direct aim at Anthropic's framing of Mythos as a cybersecurity tool so dangerous it cannot be released to the general public, likening the approach to a company saying, "We have built a bomb, we are about to drop it on your head. We will sell you a bomb shelter for $100 million." The analogy encapsulates Altman's broader critique: that Anthropic is deliberately amplifying fears about Mythos's destructive potential in order to justify restricting access and command premium positioning with elite institutional clients.
Anthropic announced Claude Mythos in April 2026 as an advanced AI system capable of autonomously identifying cybersecurity vulnerabilities — a capability the company acknowledged places it beyond the threshold of responsible public release. Rather than offering open access, Anthropic launched Project Glasswing, a restricted program granting access to just 11 organizations, including Google, Microsoft, Amazon Web Services, Nvidia, and JPMorgan Chase. The stated rationale is defensive: allowing major infrastructure operators to use Mythos to stress-test and harden their own systems. Simultaneously, Anthropic has been conducting a phased rollout of Claude Opus 4.7, described as a less advanced model, suggesting a deliberate tiering strategy in which cutting-edge capabilities are sequestered behind institutional gatekeeping while more modest versions serve broader markets.
The controversy taps into longstanding tensions within the AI industry over who controls the most powerful systems and under what justification. Altman acknowledged that legitimate AI safety concerns exist, but argued that Anthropic's rhetoric crosses into strategic exaggeration — using the specter of catastrophic risk to rationalize exclusivity rather than as a genuine safeguard. This critique resonates with a recurring debate in AI policy circles about whether safety-driven access restrictions serve the public interest or primarily entrench advantages for well-capitalized incumbents and their preferred partners. The Barclays CEO's warning that Mythos could catalyze bank cyberattacks, cited in related coverage, illustrates how seriously the financial sector is taking Anthropic's own framing — which in turn lends credence to Altman's point that the fear narrative has real-world persuasive power.
The exchange marks a notable escalation in the public rivalry between OpenAI and Anthropic, two companies that share deep ideological roots — Anthropic was founded by former OpenAI executives — but have increasingly diverged in their public positioning and commercial strategies. Anthropic has consistently leaned into a safety-first identity, while OpenAI under Altman has pushed for broader access and faster deployment cycles. The Claude Mythos episode crystallizes that divide: Anthropic is betting that controlled, high-value exclusivity is both the responsible and commercially viable path for frontier AI, while Altman is framing that bet as self-serving theater. Whether the market ultimately rewards cautious gatekeeping or expansive access will have significant implications for how both companies — and the broader industry — develop and deploy increasingly capable AI systems in the years ahead.
Read original article →