Anthropic Acquires AI Startup Vercept to Advance Claude Computer Use Capabilities
Original: Anthropic Acquires AI Startup Vercept to Advance Claude Computer Use Capabilities View original →
Vercept Acquisition Announced
Anthropic announced on February 25, 2026 via X: "Anthropic has acquired Vercept_ai to advance Claude's computer use capabilities." The acquisition comes as competition intensifies around AI agents capable of autonomously operating software.
Vercept's Technology and Team
Seattle-based Vercept, which raised $50 million total with backing from investors including former Google CEO Eric Schmidt and Google DeepMind's Jeff Dean, developed technology enabling AI to autonomously control desktop software. Co-founders Kiana Ehsani, Luca Weihs, and Ross Girshick will join Anthropic with their team.
Major Leap in Claude Computer Use
According to Anthropic, Claude Sonnet 4.6 advanced from under 15% to 72.5% on OSWorld — a leading benchmark for AI computer use — approaching human-level performance on tasks like navigating complex spreadsheets and completing web forms across browser tabs.
Product Transition
Vercept's desktop application Vy will shut down within 30 days post-acquisition, with existing users encouraged to migrate to Claude tools. The co-founders said their startup's mission had found a bigger home at Anthropic, citing the lab's focus on building safe, steerable AI systems.
Related Articles
Anthropic says it has acquired Vercept to push Claude’s computer-use capabilities further. The company also tied the move to Sonnet 4.6 progress, citing a rise to 72.5% on OSWorld.
Anthropic published a March 6, 2026 case study showing how Claude Opus 4.6 authored a working test exploit for Firefox vulnerability CVE-2026-2796. The company presents the result as an early warning about advancing model cyber capabilities, not as proof of reliable real-world offensive automation.
Anthropic published a Mar 6, 2026 policy for vulnerabilities identified with Claude. The framework sets a 90-day default disclosure window, a 7-day target for actively exploited critical bugs, and human review requirements before reports go out.
Comments (0)
No comments yet. Be the first to comment!