Here's what's new in the world of AI tools relevant to legal professionals this week.

1. A&O Shearman Receives AI Revenue Share with Harvey. A&O Shearman and Harvey AI announced on April 6, 2025, a plan to jointly develop and deploy "agentic AI agents" targeting complex legal workflows like antitrust analysis and loan review. According to the announcement, these agents will integrate A&O Shearman's legal expertise and will be offered commercially via subscription or usage fees to corporations, financial services firms, and other law firms. Notably, A&O Shearman stated it will share in the software revenue generated from these sales. This partnership model presents a different approach to integrating AI compared to strategies like Cleary Gottlieb's acquisition of AI developer Springbok AI (covered previously), representing a direct co-development and revenue-sharing arrangement between a law firm and an AI provider.

2. Google Announces AI Agent Protocol. Following OpenAI's support for the Model Context Protocol (MCP) announced March 27, 2025 (covered previously), Google launched its own complementary open standard, the Agent2Agent (A2A) protocol, on April 9, 2025. While MCP focuses on standardizing how AI agents securely access external tools and data sources, Google states A2A is designed to standardize how different AI agents (potentially from various vendors) communicate and coordinate tasks directly with each other. Together, these protocols signal a move towards greater interoperability in the rapidly developing AI agent ecosystem.

3. Google Agentspace and Email Actions. Google's Agentspace platform, designed as an enterprise hub connecting work applications and AI, includes capabilities for automating tasks. Among the actions presented this week at Google Cloud Next were integrations with common business tools to perform tasks like sending emails and scheduling meetings, as demoed this week.

4. ChatGPT Can Now Access All Past Chats. OpenAI announced yesterday (April 10, 2025), that ChatGPT's memory feature for Plus and Pro subscribers has been enhanced. The system can now automatically reference a user's entire chat history to provide more personalized and context-aware responses, moving beyond previously saved explicit memories. Users retain control and can disable the feature. Wharton Professor Ethan Mollick commented on the update, acknowledging the potential utility but also noting his personal preference for maintaining boundaries between personal context and work-related AI interactions.

5. Reports Suggest Enhanced Google Deep Research Activity. Recent user reports on social media platforms, suggested that Google's Deep Research feature (within Gemini) may be crawling significantly more web pages (potentially thousands per query) than previously observed, even for free-tier users. This anecdotal evidence points to a possible, unannounced scaling up of the feature's web-crawling capabilities.

6. Meta Releases Llama 4 with 10M Token Context. On April 5, 2025, Meta released Llama 4 Scout and Maverick, its latest open-weight, multimodal AI models built with a Mixture-of-Experts architecture. A standout feature is Llama 4 Scout's reported 10 million token context window, representing a substantial increase over previous models and enabling tasks involving much larger amounts of text or data.

Disclaimer: The author practiced law for six years at Ropes & Gray LLP prior to co-founding UseJunior. UseJunior is a member of both Google for Startups and Microsoft for Startups and has received grants and in-kind contributions in connection with these programs. The views expressed herein are our own and do not represent the views of Google, Microsoft or Ropes & Gray LLP. This is not legal advice.