NVIDIA's Big Announcement Rush at Computex 2026
At Computex 2026, currently underway in Taipei, NVIDIA announced a slate of major products all at once. The three pillars — the world foundation model **Cosmos 3**, the inference-focused LLM **Nemotron 3 Ultra**, and the real-time streaming AI platform **RTX Spark** — show that NVIDIA's AI ecosystem is extending from the cloud all the way to edge devices.
Latent Space described the announcement as "Jensen scores a huge win," with CEO Jensen Huang once again demonstrating his overwhelming presence in the AI industry.
CEO Jensen Huang: "People Should Be Paid as Much as Possible"
Also at the Computex venue, CEO Huang answered a question about employee compensation, stating that "**people should be paid as much as possible**." It is a clear stance from NVIDIA's chief on the global debate over how to distribute the enormous profits generated by the AI infrastructure boom.
The remark came in response to a question about a recently announced compensation deal at partner company Samsung that includes bonuses of up to $400,000. Whether the benefits of the AI boom flow back to workers has become a major issue for the tech industry as a whole.
Anthropic Expands General Access to the Claude Mythos Model
Anthropic has expanded general access to the AI model **Claude Mythos**. By broadening what had until now been a limited access quota, more developers and enterprises will be able to use Claude Mythos.
Anthropic is known for its safety-first corporate posture, and this access expansion appears to be a phased rollout undertaken once its risk management safeguards were in place. Further official announcements detailing the conditions and scope of use are expected to follow.
China Adds AI and Data to Its Trade Secret Rules
China has announced regulatory revisions that newly add AI and data-related assets to the scope of trade secret protection. The move is seen as intended to prevent the outflow of AI technology.
As the AI technology race between the US and China intensifies, the change signals that China is strengthening its legal framework. AI models and training data are being positioned as strategic assets, and information management requirements imposed on companies and researchers may become stricter.
JetBrains Open-Sources the Code AI Model Mellum2
JetBrains has released **Mellum2**, a model developed for AI-powered coding assistance, as open source. The model underpins JetBrains' AI assistant features and is characterized by fast code completion and workflow integration.
Open-sourcing it will allow the community at large to contribute to improving the model, and Mellum2 is also expected to find use in environments beyond JetBrains IDEs. That adds one more open source option in the coding AI space.
AWS Publishes a Guide to Implementing OAuth Authentication for MCP Servers on AgentCore Gateway
AWS has published a technical blog post explaining how to implement secure authentication, using the OAuth authorization code flow, for MCP servers hosted on Amazon Bedrock AgentCore Gateway.
The mechanism ensures that requests from AI assistants (such as the Kiro IDE) to MCP servers are authenticated with valid tokens issued by an organization's identity provider (Okta, Entra ID, Cognito, and others). It serves as a valuable, practical reference for implementing MCP server security in enterprise environments.
Other Notable Moves
ShadowProtect**: A "Wireshark for AI Agents"-style tool that captures and analyzes AI agent traffic drew attention on Hacker News.
American AI Sovereign Wealth Fund Act**: The "American AI Sovereign Wealth Fund Act" was introduced in the US Senate, kicking off debate over a national investment framework for the AI sector.
Meta AI Support System Prompt Leak**: The system prompt behind Meta's AI support feature was leaked and posted on GitHub.
---