Anthropic’s Mythos Under Fire as OpenAI Attacks Its Marketing and Unauthorised Users Breach the Model

Why It Matters

The clash underscores heightened competitive tension and raises questions about AI safety claims, while the breach highlights persistent security vulnerabilities in tightly controlled models.

Key Takeaways

• OpenAI CEO calls Anthropic’s limited release “fear‑based marketing”.
• Mythos access limited to ~40 firms under Project Glasswing.
• Unauthorized users breached Mythos via third‑party vendor exploit.
• Anthropic says no impact on its own systems yet.
• OpenAI unveiled ChatGPT Images 2.0 with web‑search and 2K output.

Pulse Analysis

Anthropic introduced Mythos as a highly restricted AI designed to bolster cybersecurity defenses, limiting access to a curated group of about 40 organizations. The company framed the restriction as a safeguard against weaponisation, positioning the model as a premium, risk‑managed service under the Project Glasswing banner. This approach reflects a broader industry trend where firms balance innovation with responsible deployment, especially as AI capabilities increasingly intersect with critical infrastructure.

Sam Altman's public dismissal of Anthropic’s rollout strategy on the Core Memory podcast amplified the competitive narrative between the two AI powerhouses. By branding the limited release as fear‑based marketing, OpenAI not only challenged Anthropic’s safety narrative but also sought to position its own offerings as more accessible and trustworthy. The subsequent Bloomberg report of unauthorized access—traced to a third‑party contractor’s credentials—adds a layer of credibility to Altman’s critique, raising concerns about the robustness of access controls even for tightly gated models.

In parallel, OpenAI’s debut of ChatGPT Images 2.0 signals a strategic pivot toward multimodal, consumer‑friendly AI products. The new model’s web‑search integration, multi‑panel comic generation, and 2 K resolution output broaden its appeal beyond developers to marketers and creators. By expanding its portfolio while questioning competitors’ safety postures, OpenAI reinforces its market dominance and pressures rivals like Anthropic to reassess both security protocols and go‑to‑market strategies. Enterprises watching the AI arms race must weigh the trade‑offs between cutting‑edge functionality and the assurance of rigorous security safeguards.