Microsoft readies cheaper AI security rival to Anthropic

Microsoft is reportedly preparing a new AI security product called Project Perception, positioning it as a lower-cost rival to Anthropic’s Mythos. According to The Information, the tool would be deployed inside an organization’s IT environment to detect vulnerabilities, then scan, identify, and suggest fixes using a mix of models from Anthropic, OpenAI, and Microsoft.

The pitch appears to be as much about price as performance. The report says Mythos carries an estimated API cost 100 percent higher than Opus and 82 percent higher than GPT, both described as among the most expensive publicly available AI models. Microsoft reportedly plans to cut costs by routing each query to a specific model based on the task, a strategy already used across the industry as companies look for cheaper ways to run AI services.

Microsoft’s enterprise AI push

The reported product fits a broader shift inside Microsoft toward enterprise-focused AI. Mustafa Suleyman, the company’s AI lead, has said he wants Microsoft to launch frontier AI models next year without relying on distillation from other models. The company is also rolling out hardware designed for AI use.

At the same time, Microsoft has sharpened its criticism of partners that are now also competitors, including OpenAI and Anthropic. CEO Satya Nadella said those companies were duping customers by using their data to improve their own models, while suggesting Microsoft had safeguards to avoid doing the same. Microsoft’s sales pitch, as the article frames it, is increasingly straightforward: better security and governance, at a lower cost.

That message matters because Microsoft already has a deep foothold in enterprise IT through Windows and Office, even as Anthropic, OpenAI, and Gemini remain better known for their AI models.

Recommended reading

Carders Want 'Clean' Residential Proxies Now

Export rules could shape the rollout

Microsoft may also face fewer export-related hurdles than Anthropic in taking an AI cybersecurity product to Europe and other regions. The article notes that Anthropic’s Fable 5 model — trained similarly to Mythos but with added guardrails — was barred from export outside the US, forcing Anthropic to suspend access while it worked with the US government on safety verification. Microsoft also restricted Fable access for Azure customers.

Even so, demand has held up. Several US agencies still use Mythos despite the US government designating Anthropic a supply risk, and governments and financial institutions in Europe have been seeking access through US channels.

For businesses, more products in this category could soon mean cheaper vulnerability detection and remediation, especially as attackers gain access to the same advanced tools. The urgency is already clear: a recent critical Zoom flaw exposed Windows users to potential data theft and system compromise before a patch was released.