This week saw significant developments in the AI landscape, including major model releases, strides in AI agent standardization, enhanced audio capabilities, and noteworthy advancements from international players. Staying informed about these rapid changes is crucial for legal professionals navigating the evolving technological environment.

1. Google Releases Experimental Gemini 2.5 & Canvas Collaboration Tool

* Google released the experimental version of its next-generation AI model, Gemini 2.5, on March 25, 2025.^[1]

* Gemini 2.5 is Google's highest-performing model to date, demonstrating leading results on complex benchmarks such as GPQA and AIME 2025, potentially offering advanced capabilities for challenging legal research and analytical tasks.

* Additionally, on March 18, 2025, Google introduced Canvas, a collaborative workspace integrated with Gemini.^[2] Canvas features real-time code editing, previews, and document creation flows, aiming to enhance productivity for tasks like collaborative legal drafting and case preparation by embedding AI assistance directly into shared workflows.

1. OpenAI Embraces Model Context Protocol (MCP) Standard

* OpenAI announced its support for the Model Context Protocol (MCP) on March 26, 2025.^[3]

* MCP, initially developed with significant input from competitor Anthropic, aims to standardize how AI agents interact with external tools and data sources. OpenAI CEO Sam Altman confirmed support is available in the Agents SDK, with plans for future integration into ChatGPT and APIs. This adoption by OpenAI signals growing momentum for MCP as a potential industry standard, which could simplify the integration of AI into various legal tech tools, such as document management systems.

3. Zapier Adds MCP Support, Signaling Wider Adoption

* The automation platform Zapier announced its support for MCP on March 24, 2025.4] This integration allows AI agents built using the standard to potentially connect with Zapier's [extensive library of over 8,000 applications and 30,000 actions without needing complex, custom API work.^[5]

* Zapier's adoption further underscores the growing industry acceptance of MCP, highlighting its potential to streamline automation workflows relevant to legal practice, such as e-discovery processes or automated client communications.

4. OpenAI Launches New API Models for Advanced Audio Agents

* On March 20, 2025, OpenAI introduced three new API-based audio models: gpt-4o-mini-tts, gpt-4o-transcribe, and gpt-4o-mini-transcribe.^[6]

* These models offer enhanced text-to-speech (TTS) and speech-to-text (STT) capabilities for building voice agents. Notably, gpt-4o-mini-tts allows for adaptive speaking styles based on prompts, while the transcription models reportedly outperform the previous Whisper model in accuracy across benchmarks.

5. Baidu Introduces New ERNIE AI Models in China

* Chinese technology firm Baidu launched two new AI models on March 16, 2025: ERNIE 4.5 and ERNIE X1.^[7]

* ERNIE 4.5 is a natively multimodal model that Baidu claims offers performance exceeding models like GPT-4.5 at a significantly lower cost (reportedly 1% of the cost, with input/output pricing from RMB 0.004/0.016 per 1k tokens). ERNIE X1 is presented as Baidu's first reasoning-focused AI, aiming to match the performance of competitors like DeepSeek's R1 at half the price.

Bonus Update:

* Ghibli-Style AI Art Captures Online Attention: Following updates to OpenAI's models (like GPT-4o), a viral trend emerged this week involving users generating images mimicking the distinctive style of Studio Ghibli animations.8] Users are sharing these creations widely and even combining them with video tools like [Runway to create short animations.^[9] While demonstrating AI's increasing mainstream cultural impact and creative accessibility, this trend has also sparked discussions around potential copyright infringement and the unauthorized exploitation of artistic styles, raising relevant considerations for intellectual property law practitioners.

Conflicts of Interest. 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.

References:

1. Google, Gemini model thinking updates: March 2025, Google AI Blog (Mar. 25, 2025), https://blog.google/technology/google-deepmind/gemini-model-thinking-updates-march-2025/

2. Google, Gemini Canvas Overview, Gemini (Mar. 18, 2025), https://gemini.google/overview/canvas/

3. Kyle Wiggers, OpenAI adopts rival Anthropic's standard for connecting AI models to data, TechCrunch (Mar. 26, 2025), https://techcrunch.com/2025/03/26/openai-adopts-rival-anthropics-standard-for-connecting-ai-models-to-data/

4. Rafal Wilinski, X Post, X (Mar. 24, 2025), https://x.com/rafalwilinski/status/1903811725360591087

5. Zapier, Model Context Protocol (MCP), Zapier (Mar. 24, 2025), https://zapier.com/mcp

6. OpenAI, New audio models in the API: tools for voice agents, OpenAI Developer Community (Mar. 20, 2025), https://community.openai.com/t/new-audio-models-in-the-API-tools-for-voice-agents/1148339

7. Baidu, Baidu Unveils ERNIE 4.5 and Reasoning Model ERNIE X1 – Makes ERNIE Bot Free Ahead of Schedule, PR Newswire (Mar. 16, 2025), https://www.prnewswire.com/news-releases/baidu-unveils-ernie-4-5-and-reasoning-model-ernie-x1--makes-ernie-bot-free-ahead-of-schedule-302402490.html

8. Variety, OpenAI CEO on Viral Studio Ghibli-Style AI Art: 'We're Not Trying to Replace Artists', Variety (Mar. 27, 2025), https://variety.com/2025/digital/news/openai-ceo-chatgpt-studio-ghibli-ai-images-1236349141/

9. Runway, Runway, Runway (last visited Mar. 28, 2025), https://runwayml.com/