OpenAI launches ChatGPT Gov while looking to raise $40Bn + Deepseek continues to disrupt AI industry

Today’s Highlights:

?? News: OpenAI launches ChatGPT Gov and Deepseek faces major scrutiny and bans

?? Funding: OpenAI looks to raise $40Bn

?? Top News Stories:

1. OpenAI launches ChatGPT Gov, providing U.S. agencies a secure AI platform within Microsoft Azure, providing GPT-4o-powered tools while meeting strict federal security standards.

• ChatGPT Gov is a secure AI platform designed specifically for U.S. government agencies, running on Microsoft Azure’s commercial and government clouds, compliant with IL5, CJIS, ITAR, and FedRAMP High security standards.
• The platform grants agencies access to GPT-4o and AI tools for text interpretation, summarization, coding, image analysis, and mathematics, with features for secure conversation sharing and file uploads.
• Agencies can build and share custom GPTs, while IT administrators have centralized controls to manage users, AI models, and security settings, ensuring streamlined AI governance.
• ChatGPT Gov adoption is already widespread, with over 90,000 government users across 3,500 agencies exchanging 18 million messages since 2024, including use by the Air Force Research Laboratory.

2. Deepseek’s launch has continued to send waves throughout the AI + Tech industry:

• Microsoft is investigating claims that DeepSeek used OpenAI’s API to train its own AI models, potentially violating OpenAI’s terms of service and raising legal concerns.
• Microsoft has swiftly integrated DeepSeek’s R1 AI model into Azure AI Foundry and GitHub, betting on its compute efficiency to drive down AI costs and expand business applications.
• DeepSeek exposed a database containing secret keys, chat logs, and operational data, raising serious security concerns despite quickly patching the vulnerability.
• Hundreds of companies and government agencies have restricted access to DeepSeek’s AI tools over concerns about data leaks to the Chinese government, weak privacy safeguards, and cybersecurity risks, with Italy’s privacy regulator blocking DeepSeek entirely and Ireland launching an investigation into its data practices.
• The U.S. Navy has prohibited the use of DeepSeek over security and ethical concerns, warning personnel not to use it for any purpose.
• The Pentagon is rushing to block DeepSeek after Department of Defense employees accessed the Chinese AI service on work computers for at least two days, raising national security concerns due to its data storage under Chinese jurisdiction.
• India plans to host DeepSeek’s AI models on domestic servers under strict data localization rules, signaling a rare embrace of Chinese technology amid India’s growing AI ambitions.
• DeepSeek’s AI refuses to answer 85% of politically sensitive prompts related to China, often substituting responses with nationalistic messaging, according to a new report.
• David Sacks, Trump’s AI and crypto advisor, claims there is “substantial evidence” that DeepSeek used OpenAI’s models to train its own, comparing the alleged practice to theft through distillation.
• Anthropic CEO Dario Amodei asserts that DeepSeek’s AI progress confirms that U.S. export controls are effective, as Chinese models still lag behind leading U.S. models despite cost efficiencies.
• Hugging Face researchers have launched Open-R1, an initiative to fully replicate DeepSeek’s R1 AI model and openly share its training data, methodology, and architecture.

3. Block, Jack Dorsey’s fintech company, has launched Goose, an open-source AI agent framework that enables developers to build and deploy AI assistants across multiple platforms while maintaining data privacy and control through local execution.

4. Figure AI is launching the Center for the Advancement of Humanoid Safety, a new initiative aimed at setting industry-leading safety standards and transparency in AI-controlled humanoid robotics, beginning with their Figure 02 model.

5. OpenAI, in collaboration with Microsoft, is providing AI models to U.S. National Laboratories for nuclear weapons security and scientific research, deploying the technology on Los Alamos’?

supercomputers.

6. The U.S. Copyright Office (USCO) ruled that AI-generated works without human modification cannot receive copyright protection, while creations involving meaningful human authorship can be copyrighted.

7. Google has officially rolled out Gemini 2.0 Flash as the default AI model in the Gemini app, delivering faster responses and stronger performance across key benchmarks.

8. Mistral AI has launched Mistral Small 3, a 24-billion-parameter open-source language model that offers performance comparable to larger models like Llama 3.3 70B while being faster and more cost-efficient.

?? Top Funding News:

1. OpenAI is in early talks to raise up to $40Bn, led by SoftBank, which may invest between $15Bn and $25Bn, potentially valuing the ChatGPT maker at $300Bn.

• The funding will help OpenAI finance its massive $18Bn commitment to Stargate, a joint venture with SoftBank and Oracle to build new U.S. data centers for AI infrastructure.
• If successful, OpenAI’s valuation would double from its October 2024 figure of $157Bn, making it the second-most valuable startup in the world, behind only Elon Musk’s SpaceX.

2. Skild AI, a robotics AI company building a foundational model for adaptable automation, is in talks to secure a $500M investment from SoftBank at a $4B valuation, following previous backing from Jeff Bezos, Lightspeed Venture Partners, and Coatue.

3. Quibim, an AI-driven medical imaging startup improving disease detection and diagnostics, raised a $50M Series A led by Asabys and Buenavista.

4. Atomicwork, an AI-driven IT service management platform automating enterprise workflows, raised a $25M Series A co-led by Khosla Ventures and Z47, with participation from Battery Ventures, Blume Ventures, and Peak XV Partners.