Anthropic launches Claude Opus 4.7 as Mythos raises legal AI stakes
Anthropic released Claude Opus 4.7 on April 16, 2026, and previewed the more advanced Mythos model.
Why it matters: Legal teams rely on foundational AI models for contract review, compliance, and data handling. Anthropic’s latest launches—and the implications of Mythos’ vulnerability-finding abilities—raise new regulatory and operational stakes for legal tech adopters.
- Claude Opus 4.7 debuted April 16, 2026, offering upgraded coding, vision, and self-critique features.
- Anthropic confirmed Opus 4.7's capabilities are surpassed by Mythos, its upcoming, restricted-access model.
- Mythos has identified thousands of critical software vulnerabilities, prompting tight expert-only access.
- Claude Cowork for enterprise adds automation, analytics, and access controls for legal workflows.
Anthropic launched Claude Opus 4.7 on April 16, 2026, delivering notable upgrades for legal and business users: improved coding assistance, expanded multimodal (vision) features, and enhanced self-checking to flag potential errors or hallucinations. The company says these enrichments are part of its broader push to offer more reliable tools for enterprise and legal tech solutions.
- While announcing Opus 4.7, Anthropic confirmed that its internal benchmarks place the new model behind Mythos, its forthcoming model under Project Glasswing. CEO Dario Amodei stated, “Opus 4.7 is excellent, but Mythos is larger, more intelligent, and represents a leap ahead.”
- Mythos attracted intense industry scrutiny when, during preview tests, it discovered thousands of zero-day software vulnerabilities in widely used operating systems and browsers. The model’s advanced reasoning raises both security opportunities and legal risks, as widespread unpatched bugs expose enterprises to compliance and cyber threat concerns.
- Due to the profound cybersecurity implications, Anthropic is limiting access to Mythos for now to a vetted group of specialists and select enterprise partners like Amazon and JPMorgan Chase (Tom's Hardware).
- For enterprise clients—including legal departments—Claude Cowork introduces task automation, granular access controls, and analytics, supporting operational security and compliance in document-intensive environments.
Anthropic is also retiring Claude Sonnet 4 and Opus 4 on June 15, 2026, and opened public beta for Claude Managed Agents—workflows enabling sandboxed automation, which could help legal ops teams safely integrate AI-driven routines (Axios).
The dual track of rapid advancement and heightened access restrictions illustrates new challenges for legal tech strategists, especially regarding model transparency, governance, and regulatory liability.
By the numbers:
- April 16, 2026 — Date of Claude Opus 4.7 public launch
- Thousands — Number of zero-day vulnerabilities Mythos flagged in test deployments
- June 15, 2026 — Deadline for retiring older Claude models
Yes, but: Mythos' capabilities remain unproven in real-world legal scenarios, and restricted access limits broad legal evaluation.
What's next: Anthropic will fully retire Claude Sonnet 4 and Opus 4 models on June 15, 2026; access to Mythos may expand after current security reviews.