Firms Pivot to Anthropic Opus 4.7 as Mythos Remains Restricted

Organisations are increasingly deploying Anthropic's Claude Opus 4.7 to address cybersecurity vulnerabilities, as the more powerful Mythos model remains largely inaccessible to the broader market. The shift represents a pragmatic recalibration in enterprise AI adoption, where deployability is trumping raw capability.

According to Sangeeta Gupta, chief strategy officer of Nasscom, Opus 4.7 offers approximately 70-80% of Mythos capabilities, making it a practical interim solution for threat hunting and incident response workflows. The model's accessibility has positioned it as the default choice for enterprises that cannot wait for broader Mythos access.

Anthropic's official blog post confirms the company deliberately reduced Opus 4.7's cybersecurity capabilities during training compared to Mythos. The firm introduced safeguards that automatically detect and block requests linked to prohibited or high-risk cybersecurity use cases. This approach allows Anthropic to test cyber safeguards on less capable models before any broader Mythos release.

The performance gap is measurable. Mythos scored 83.1% on CyberGym, a cybersecurity capability benchmark developed by researchers at the University of California, Berkeley, while Opus 4.7 scored 73.1%. For most enterprise use cases, that 10-point difference barely registers in daily operations (though security teams will notice the edge cases).

Jeremy D'Hoinne, vice president analyst at global research firm Gartner, noted that the differences between Opus 4.7 and Mythos would only be visible on some edge cases. He observed that enterprises are now running experiments with models they currently have access to, while cybersecurity vendors have started integrating AI-driven scanning and patching capabilities into their products.

The pricing structure remains unchanged from Opus 4.6: $5 per million input tokens and $25 per million output tokens. Developers can access the model via the Claude API, Amazon Bedrock, Google Cloud's Vertex AI, and Microsoft Foundry. This multi-platform availability reduces friction for organisations already invested in these ecosystems.

For many enterprises, especially in regulated sectors, models like Claude Opus 4.7 are proving sufficiently capable for threat hunting, anomaly detection, compliance automation, and incident response workflows. The physical reality of using these systems matters: developers report being able to hand off their hardest coding work to Opus 4.7 with confidence, catching logical faults during the planning phase rather than debugging after deployment.

Arjun Nagulapally, chief technology officer at AIONOS, an AI operating system delivery firm, stated that within 18 months this capability will become table stakes for enterprise AI. He emphasised that companies moving fastest aren't chasing maximum capability but rather minimum friction with sufficient capability.

Most enterprises do not necessarily require frontier-class AI systems. Instead, they are looking for models that are powerful enough to handle complex cybersecurity and automation tasks while remaining easier to deploy across existing operations. The latency improvements and adaptive thinking features mean engineers can stay in flow rather than waiting for model responses.

Advanced AI systems are already changing how enterprises handle security operations across cloud, network, endpoint, and identity systems. Beyond operational workflows, experts warned that AI systems are also expanding the attack surface on older infrastructure that was never built to withstand this level of automated scrutiny.

Jaydeep Singh, general manager at cybersecurity solutions firm Kaspersky, noted that core banking infrastructure, payment rails, and grid controllers built in the 1980s and 1990s were protected less by formal security properties than by the sheer cost of effort required to probe them. AI systems are rapidly eroding that protection.

The Mythos situation remains complicated. Fortune reported that Anthropic acknowledged training and testing a new model following a data leak that revealed its existence. The company described Mythos as representing a step change in AI performance and the most capable model it has built to date.

Anthropic is being deliberate about how it releases Mythos, working with a small group of early access customers to test the model. The company expressed concern that the model poses unprecedented cybersecurity risks, with capabilities that could enable large-scale cyberattacks if misused.

The World Economic Forum highlighted that Anthropic's decision to limit access to Mythos reflects a growing focus on safe, secure, and responsible deployment of advanced AI. The cautious rollout strategy underscores a new reality where AI capability is advancing faster than the ability to safely govern it.

Industry experts said the shift signals a broader change in enterprise AI where businesses are increasingly prioritising deployability, workflow integration, and operational outcomes over access to the most advanced AI systems available. The fact that enterprises are evaluating capable but accessible models rather than waiting for restricted frontier-grade ones only accelerates how quickly organisations must adapt.

Whether Anthropic's restraint on Mythos pays off in the long run remains to be seen. For now, enterprises are making do with Opus 4.7, which is good enough for most tasks but leaves security teams wondering what they're missing out on. The real question isn't whether Mythos will eventually launch, but whether anyone will be able to afford it when they do.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn