Meta and Microsoft Restrict Employee Access to Anthropic Claude AI

Key Takeaways

• Meta and Microsoft are scaling back internal employee reliance on Anthropic’s Claude AI.
• The move aims to reduce internal AI spending and promote proprietary coding assistance tools.
• Microsoft has significantly cut internal monthly AI spending limits per employee, while Meta has seen a substantial drop in Claude Code users.
• These internal policy changes do not impact external customer access to Claude through Microsoft’s platforms.

Major tech giants Meta and Microsoft are implementing new policies to curb their employees’ use of Anthropic’s Claude artificial intelligence, instead encouraging the adoption of their in-house developed coding tools. This strategic pivot, initially reported by The Information on October 5, focuses on internal cost management and optimizing developer workflows. It’s crucial to note that these internal adjustments do not affect customer access to Claude through Microsoft’s existing product offerings.

This evolving landscape underscores the increasing financial pressures associated with integrating AI coding assistants into daily software development. While both companies maintain their status as significant Anthropic clients, they are also direct competitors in the AI space. Their decisions highlight the intricate connections between AI expenditure, product development strategies, and the imperative to control core developer tools.

Microsoft Cuts AI Spending Limits

Microsoft had initially projected its internal spending on Anthropic’s technology to exceed $1 billion annually. However, this forecast has reportedly been slashed by over a third. This reduction follows directives from leadership for employees to decrease their reliance on Claude and instead utilize Microsoft’s own solutions, such as GitHub Copilot and models from OpenAI.

Within Microsoft’s influential cloud and AI division, the monthly caps for AI spending have reportedly plummeted from $100,000 per employee to approximately $10,000 in most instances. These figures represent maximum spending allowances, not the actual average expenditure per employee. The introduction of stricter limits has reportedly caused some discontent among engineers who previously enjoyed greater autonomy in experimenting with various AI models.

It is important to differentiate between internal budget restrictions and external product availability. The reduced internal budgets do not signify the removal of Claude from Microsoft’s customer-facing products. Reports suggest that customer spending on Anthropic models via Microsoft’s enterprise platforms continues to expand, even as the company tightens its internal reins on employee usage.

Meta Pushes Its Coding Tools

At Meta, the number of employees using Claude Code has reportedly halved, dropping from approximately 60,000 earlier this year to around 30,000. While recent layoffs contributed to this decrease, the primary driver is Meta’s concerted effort to promote its proprietary AI tools, including MetaCode and Muse Code, which are powered by Meta’s own AI models.

MetaCode has reportedly garnered over 30,000 internal users, with Muse Code attracting more than 6,000 employee users. Meta commenced external customer trials for Muse Code in August. Despite the decline in Claude Code users, Meta reportedly incurred expenses exceeding $105 million on the tool during a recent 28-day period, indicating that a smaller user base does not necessarily equate to lower overall spending.

For cybersecurity professionals, the transition between coding assistants does not eliminate inherent risks associated with automated access to files, commands, and credentials. Previous cybersecurity reports have detailed vulnerabilities within Claude Code, specifically concerning repository settings that could have facilitated unauthorized code execution and API key exfiltration. Anthropic addressed and patched these issues prior to public disclosure.

Microsoft’s AI tools have also encountered similar security challenges. The patched RoguePilot vulnerability, for instance, demonstrated how malicious instructions embedded within a GitHub Issue could lead to a repository takeover. These incidents underscore the critical need for granular permissions, secure project configurations, and rigorous review processes for all AI-generated actions.

Anthropic’s reported annualized revenue pace of $65 billion highlights the ongoing strong demand for its AI technology. Therefore, the strategic adjustments by Meta and Microsoft reflect a focus on budget optimization and competitive positioning, rather than a complete disengagement from Claude.

Disclaimer: HackersRadar reports on cybersecurity threats and incidents for informational and awareness purposes only. We do not engage in hacking activities, data exfiltration, or the hosting or distribution of stolen or leaked information. All content is based on publicly available sources.