ChatGPT Can Now Connect to Your Apps 🤖!

Also: Mistral releases its coding agent, Mistral Code, while Anthropic announces models for US government use 🤫.

Author

The Neural Frontier
June 06, 2025

Source: ChatGPT Image Generator

There’s a new vibe-coding agent on the block, and it’s from your favorite French AI startup. 

Hello, and welcome to issue #110 of the Neural Frontier 👋. 

Today, we’re diving into some impressive ChatGPT updates, Mistral’s new coding agent, and Anthropic’s new models for military and intelligence use. 

A mixed bag? Definitely. Let’s unpack 😉! 

In a rush? Here's your quick byte: 

🤖 ChatGPT can now connect to your apps!

🧑‍💻 Mistral releases its coding agent, Mistral Code.

🤫 Anthropic announces models for US government use!

🎭 AI Reimagines: Some of the wildest examples from Mirage Studio

🎯 Everything else you missed this week.  

⚡ The Neural Frontier’s weekly spotlight: 3 AI tools making the rounds this week.

Source: OpenAI

OpenAI has just introduced powerful new integrations that allow ChatGPT to seamlessly access your documents, meetings, and files from various cloud services like Google Drive, Dropbox, Box, OneDrive, and SharePoint. 

This makes ChatGPT even more useful as an everyday business assistant, directly competing with leading productivity platforms.

Here’s what you need to know: 

📁 Connect your documents and files directly: ChatGPT’s new connectors allow you to securely access, search, and analyze files stored in popular cloud services directly from within ChatGPT. Key capabilities include:

  • Cloud Integrations: Easily retrieve and reference files across:

    • Google Drive

    • Dropbox

    • Box

    • Microsoft OneDrive and SharePoint

  • Real-time Insights: Quickly ask questions like:

    • “What were last quarter’s sales numbers?”

    • “Summarize key points from the project document.”

  • Respect for Security: ChatGPT strictly follows your organization's existing access permissions and security controls.

🎙️ Record, transcribe, and manage meetings seamlessly: ChatGPT now supports built-in meeting recording and transcription:

  • Automated Meeting Notes: Get detailed, timestamped notes with suggested action items generated automatically.

  • Interactive Queries: Instantly ask follow-up questions about your meeting’s content, transcripts, and summaries.

  • Integrated Action Items: Quickly convert meeting tasks into actionable Canvas documents for streamlined follow-up.

This feature positions ChatGPT alongside competitors like Zoom, ClickUp, and Notion, all of which offer similar transcription and summarization features.

🔎 Deep Research powered by HubSpot, Linear, and more: ChatGPT’s Deep Research now connects to specialized tools and platforms, helping you conduct comprehensive analysis by merging external data with ChatGPT’s insights:

  • Advanced Connectors: HubSpot, Linear, select Microsoft and Google services, plus web data, enabling detailed research reports with accurate citations.

  • Model Context Protocol (MCP): Pro, Team, and Enterprise users can leverage MCP to create custom connectors for personalized data integration.

💡 Availability

  • Meeting recording and transcription: ChatGPT Team, Enterprise, and Edu subscribers.

  • Cloud storage connectors: ChatGPT Team, Enterprise, and Edu users.

  • Deep Research & MCP connectors: Available to Pro, Team, and Enterprise users.

You can activate these connectors right now in your ChatGPT settings to start leveraging these powerful new productivity capabilities immediately.

Source: Mistral AI

French AI startup Mistral has just launched Mistral Code, a powerful new AI-driven coding assistant designed for enterprise developers. This release marks Mistral's entry into the rapidly expanding market of AI coding tools, where it aims to compete directly against industry leaders like GitHub Copilot, Windsurf, and Anysphere’s Cursor.

Here’s the lowdown: 

🛠️ What is Mistral Code? Mistral Code is a sophisticated coding client built upon the open-source project Continue, enhanced specifically for enterprise use. It integrates AI models, coding assistance, local deployment flexibility, and robust administrative tools into one fully supported package.

  • Coding Assistant: Provides real-time completions, multi-step refactoring, and in-depth assistance directly within your coding environment.

  • Supported IDEs: Available now in private beta for JetBrains IDEs and VS Code.

  • Deployment Options: Supports cloud deployment, reserved capacity, and secure, air-gapped, on-prem GPUs.

🌟 Key Features & Capabilities: Mistral Code leverages multiple state-of-the-art AI models:

  • Codestral: For rapid code completion and fill-in-the-middle suggestions.

  • Codestral Embed: Enhances code search and retrieval.

  • Devstral: Handles more complex, "agentic" coding tasks.

  • Mistral Medium: Powers intuitive chat assistance and developer interaction.

Mistral Code also offers:

  • Support for over 80 programming languages.

  • Integration with third-party plugins.

  • Comprehensive code reasoning across files, Git diffs, terminal output, and issue tickets.

🔒 Enterprise-Ready with Security & Compliance: Mistral specifically designed its coding assistant to address key enterprise pain points identified through direct feedback from engineering leaders and CISOs:

  • Secure & Private: All models and code reside securely within the enterprise boundary, avoiding external exposure.

  • Customizable: Organizations can fine-tune models using their own private repositories and internal data, creating lightweight, custom variants.

  • Granular Admin Console: Detailed observability, user seat management, advanced RBAC, audit logging, and usage analytics give IT teams full control and transparency.

🏢 Already Trusted by Major Enterprises: Several major enterprises have adopted Mistral Code:

  • Capgemini: Deploying on-premises for 1,500+ developers working in regulated industries.

  • Abanca Bank (Spain and Portugal): Using a hybrid deployment to securely handle sensitive banking code.

  • SNCF (French national railway): Empowering over 4,000 developers through Mistral's serverless implementation.

🤝 Open-Source Commitment: Mistral Code remains committed to the open-source community, building upon and contributing back to the original Continue project. While enhancements like multi-line editing, improved chat, and enterprise-grade functionality have been added, Mistral intends to share significant future improvements back to the community.

🌐 Availability & Getting Started

  • Currently available in private beta for enterprise users.

  • General availability is planned soon, with serverless, cloud, and self-hosted deployment options.

Interested organizations can request beta access through their Mistral account team and start enhancing their coding workflows.

Source: Anthropic

 Anthropic has officially unveiled Claude Gov, a new series of specialized AI models designed explicitly for U.S. national security and defense agencies. This release places Anthropic in direct competition with OpenAI’s ChatGPT Gov and other similar government-focused AI offerings.

As always, here’s the 411: 

🔎 What is Claude Gov? Claude Gov is a set of advanced AI models tailored specifically to the requirements of national security and intelligence agencies. These models were built with input from government partners to address real-world operational demands such as:

  • Intelligence Analysis

  • Strategic Planning

  • Operational Support

  • Cybersecurity Intelligence

The new models provide tailored capabilities, enabling them to:

  • Engage more freely and directly with classified material.

  • Demonstrate deeper understanding of context in defense and intelligence documents.

  • Offer enhanced language proficiency in dialects relevant to national security operations.

  • Accurately interpret complex cybersecurity data to inform intelligence activities.

📌 Key Differentiators for National Security: While Claude Gov models maintain the rigorous safety standards of Anthropic’s consumer and enterprise models, they have some distinct adaptations:

  • Reduced Guardrails: Unlike consumer models, Claude Gov is designed to "refuse less" when dealing with classified information, facilitating smoother integration into sensitive intelligence workflows.

  • Advanced Context Understanding: Tailored specifically for national security, these models have enhanced proficiency in interpreting nuanced and specialized government documents and data.

  • Language Specialization: They support improved language capabilities tailored for national security scenarios, helping agencies to interpret intelligence from varied linguistic sources.

🚧 Access and Usage Policies

  • Restricted Access: Claude Gov models are exclusively available to authorized U.S. national security and defense agencies operating within classified environments.

  • Carefully Managed Usage: Anthropic’s standard policies prohibit harmful activities such as weapon production, malicious cyber operations, disinformation campaigns, and censorship. However, certain usage restrictions have been adjusted to accommodate legitimate and beneficial governmental uses, with stringent oversight.

Government use of AI is closely watched due to past issues involving bias, wrongful arrests, predictive policing controversies, and discriminatory impacts. Anthropic emphasizes balancing beneficial use with harm mitigation, clearly outlining permitted activities and firmly prohibiting malicious use cases.

Source: Hamza Khalid via X

Captions AI recently launched its video model, Mirage Studios, and the reception has been pretty great. So much so that many have already started experimenting with influencer-style videos, street interviews, and a couple of other forms. 

But is it really all that? Only one way to find out!

🎯 Everything else you missed this week. 

Source: Agility Robotics 

⚡ The Neural Frontier’s weekly spotlight: 3 AI tools making the rounds this week. 

1. 📸 Picbolt: Picbolt is a browser-based visual design toolkit that transforms ordinary screenshots into professional-looking graphics without requiring design skills. 

2. 🌐 AddSubtitle: AddSubtitle is an AI-powered video editing platform that automatically generates and translates subtitles across 100+ languages with high accuracy. 

3. 📋 Writegenic AI: Writegenic AI is an industrial-grade AI writing platform that generates full-length business and project management documents in minutes.

Wrapping up… 

If this issue has taught us anything, it’s that AI is expanding faster than ever—from ChatGPT integrating into your favorite apps to Mistral redefining coding assistants and Anthropic tailoring models for government use. 

As always, we’ll keep scouring so you stay on top of every breakthrough. If you’ve enjoyed today’s insights, pass this along to a friend who needs to see what’s next in AI, and be sure to hit subscribe if you haven’t already. 

Until next week—stay curious! 😊