🤖 The Blockbrain Brief #19

Anthropic expands enterprise capabilities while defending its IP, AI agents become official team members in Jira, and Mistral partners with Accenture to scale AI adoption.

📩 High-level Summary

• The Agentic Ecosystem: Trace secured $3M to map out enterprise workflows for AI, Atlassian integrated digital agents directly into Jira to work alongside humans, and Anthropic acquired Vercept to advance Claude's ability to navigate everyday software interfaces.
• Platform & Tool Integrations: Anthropic rolled out private, customizable plugin marketplaces for Claude Enterprise, while OpenAI’s Codex partnered with Figma via MCP to let developers generate running UIs directly from code.
• Enterprise Security & Governance: Nexos.ai raised €30M to combat fragmented "Shadow AI" with a unified workspace, and Anthropic launched Claude Code Security to act as an AI-driven researcher hunting for hidden software vulnerabilities.
• Strategic Partnerships & IP Protection: Mistral AI inked a multiyear deal with Accenture to co-develop enterprise solutions, and Anthropic uncovered a massive, coordinated data distillation campaign by rival AI labs DeepSeek, Moonshot, and MiniMax.

Trace raises $3M to solve the AI agent adoption problem in enterprise

Trace, a Y Combinator-backed startup, recently secured $3 million in seed funding to address the slow adoption of AI agents within enterprises. The company specializes in workflow orchestration by building comprehensive knowledge graphs from an organization's existing tools, such as email, Slack, and Airtable. This infrastructure provides AI agents with the necessary context to scale and operate effectively across various business functions.

By mapping out these enterprise processes, Trace essentially "builds the manager" for AI labs' highly capable but context-lacking "brilliant interns." This approach ensures that AI agents can seamlessly integrate into complex corporate environments, reducing friction and enhancing overall productivity.

Jira Integrates AI Agents into the Team

Atlassian has rolled out a major update for Jira, introducing "agents in Jira." This new feature empowers users to assign, track, and manage tasks given to digital AI agents directly alongside human teammates within the same standard Jira dashboard.

This integration marks a significant shift in project management, normalizing AI agents as active participants in the workforce. By treating AI as collaborative team members rather than just background software, Atlassian aims to streamline workflows and improve visibility into both human and machine-driven tasks.

Anthropic Acquires Vercept

Anthropic has acquired Vercept, an AI startup focused on advanced software interaction, to significantly boost Claude's "computer use" capabilities. Vercept specializes in solving hard perception and interaction problems, enabling AI systems to see and act within the same software environments that humans use every day.

This acquisition aligns with Anthropic's push to move Claude beyond simple text generation toward taking on complex, multi-step tasks in live applications. With Vercept's expertise, Claude's ability to navigate interfaces, fill out forms, and manage workflows across multiple browser tabs is expected to rapidly approach human-level performance.

Anthropic upgrades Cowork and plugins on Claude for Enterprise

Anthropic has introduced a major update to Claude Enterprise, expanding its Cowork feature and rolling out an advanced plugin system. The update allows organization admins to build private, company-specific plugin marketplaces and distribute internal tools securely across various teams.

Plugin creation and management are now streamlined into a single "Customize" area, which conveniently combines plugins, skills, and Model Context Protocol (MCP) connectors under one menu. This upgrade aims to foster governed, department-level AI adoption while maintaining strict administrative controls.

Codex x Figma: Building frontend UIs

OpenAI has partnered with Figma to integrate its Codex coding assistant, allowing developers to generate Figma design files directly from Codex using the Figma MCP (Model Context Protocol) server.

This powerful integration bridges the gap between design and development, enabling users to bring real, running interfaces into Figma, refine them, and push the changes back to Codex. It represents a significant leap in automating frontend UI creation, streamlining the design-to-code pipeline for development teams.

Mistral AI Inks Strategic Pact with Accenture

Mistral AI and global consulting leader Accenture have announced a multiyear strategic partnership to co-develop enterprise technology powered by Mistral's AI models. As part of the agreement, Accenture will also become a customer, deploying Mistral's technology internally and equipping its workforce with the Mistral AI Studio.

This collaboration is designed to help organizations rapidly scale secure, enterprise-grade AI solutions tailored to specific industry challenges. By combining Mistral's high-performance open-source models with Accenture's massive global reach and industry expertise, the partnership aims to accelerate AI reinvention for clients worldwide.

Nexos.ai Raises €30M for Secure AI Governance

Lithuania's nexos.ai has raised a €30 million Series A funding round to scale its unified AI Workspace and developer Gateway. The startup aims to tackle the growing enterprise challenge of "Shadow AI": where employees use unsanctioned, fragmented AI tools, risking massive corporate data leaks.

Nexos.ai provides a secure, all-in-one platform that replaces scattered AI tools with centralized guardrails, comprehensive visibility, and flexible access controls. By offering a unified interface to manage over 200 AI models safely, nexos.ai empowers companies to adopt AI at scale without compromising on governance, security, or cost control.

Claude Code launches Claude Code Security to spot software vulnerabilities

Anthropic has unveiled Claude Code Security, a newly integrated feature within its Claude Code tool, currently available in a limited research preview. Powered by the Claude Opus 4.6 model, this AI-driven tool acts like a human security researcher, tracing data flows and understanding component interactions to catch complex vulnerabilities.

Aimed at enterprise security teams and open-source maintainers, Claude Code Security not only detects critical flaws but also suggests ready-to-apply patches for human review. The tool's powerful capabilities have already uncovered over 500 hidden flaws in production open-source projects, causing significant ripples across the cybersecurity market.

Anthropic reports distillation from DeepSeek, Moonshot, MiniMax

Anthropic recently uncovered large-scale, coordinated distillation campaigns by AI competitors DeepSeek, Moonshot, and MiniMax. These companies reportedly used over 24,000 fraudulent accounts and generated more than 16 million exchanges to secretly mine Claude’s advanced reasoning, coding, and agentic behaviors.

The sophisticated extraction efforts were verified by Anthropic through a combination of IP tracking, account metadata analysis, and API-usage patterns. In response, Anthropic has heavily updated its detection systems and tightened API controls to prevent further intellectual property theft and misuse.