NerdNews

The White House ordered Anthropic to revoke SK Telecom's access to its Claude Mythos AI model due to alleged ties to China. This led to Anthropic disabling access to the model entirely, as revoking access based on nationality would be difficult to implement while preserving privacy.

Epic Games is embracing generative AI in Unreal Engine, introducing an experimental Model Context Protocol plugin to integrate AI models like Claude and Gemini. This aims to automate tasks, increase productivity, and enhance creativity. The company plans to make the MCP an integral part of Unreal Engine 6, which is set to release in early access in late 2027.

Anthropic's AI models, Claude Fable 5 and Mythos 5, have advanced capabilities for finding software vulnerabilities and exploiting them, posing a national security risk. Experts say similar capabilities will be developed by other companies and open-source developers, making it essential for governments to develop plans for contending with AI advancements.

Chi-Hua Chien, co-founder of Goodwater Capital, believes the biggest winners of the AI era won't be companies selling AI, but rather those using AI to provide hyper-personalized experiences. He thinks infrastructure companies will get commoditized, and application companies will capture most of the value. Chien also discusses the potential of AI in enabling real-world experiences and the challenges of building a super app that combines financial services and social entertainment.

Pinterest launches 'Ask Pinterest', an experimental AI shopping app that uses conversational interface for personalized product recommendations. The app leverages Pinterest's 'Taste Graph' and can access users' saved Pins and Boards for more accurate results.

ChatGPT now allows users to schedule tasks, including monitoring tasks, with a new Scheduled page that provides a centralized location to manage upcoming requests. The feature is rolling out to Plus, Pro, Business, and Enterprise customers, with improvements to task speed and reliability.

Anthropic's design assistant, Claude Design, now integrates better with its coding agent, Claude Code. The update allows for more seamless collaboration between the two tools, enabling designers to generate assets and hand them off to developers without needing to start from scratch. The update also includes quality of life improvements, such as a more flexible import tool and a refined image editor.

Google's new $99.99 smart speaker features Gemini AI, allowing for natural language requests and multistep commands. It also includes a 'Continued Conversation' feature and ships with 10 new voices for two-way conversations. A Google Home Premium subscription plan is available for $10/month for advanced AI features.

Aditya Kumarakrishnan discusses building modular agent frameworks using CoALA, leveraging process science for scalable workflows, and 'Terraforming' legacy environments for robust agent systems. He emphasizes embracing a stronger notion of agents, building modular and extensible agents, learning from process science, and reshaping the environment for agents.

GitHub has introduced the GitHub Copilot app, a desktop control centre for agent-native development. The app aims to keep engineers in charge while AI agents handle more coding work, providing a 'My Work' view, parallel worktrees, and canvases for visible agent work. It also offers local and cloud sandboxes for contained agent execution and integrates with GitHub's pull request mechanism.

Epic Games is embracing generative AI in Unreal Engine, introducing an experimental Model Context Protocol plugin to integrate AI models like Claude and Gemini. This aims to automate tasks, increase productivity, and enhance creativity. The company plans to make the MCP an integral part of Unreal Engine 6, which is set to release in early access in late 2027.

Google's new $99.99 smart speaker features Gemini AI, allowing for natural language requests and multistep commands. It also includes a 'Continued Conversation' feature and ships with 10 new voices for two-way conversations. A Google Home Premium subscription plan is available for $10/month for advanced AI features.

Pinterest launches 'Ask Pinterest', an experimental AI shopping app that uses conversational interface for personalized product recommendations. The app leverages Pinterest's 'Taste Graph' and can access users' saved Pins and Boards for more accurate results.

Uber plans to launch a premium robotaxi service in Houston by mid-2027, in partnership with Lucid and Nuro. The service will feature Lucid Gravity SUVs equipped with Nuro's self-driving system and will compete with Waymo's commercial robotaxi services in the city.

Google's new smart speaker, available for preorder today and hitting shelves on June 25 for $99, features 360-degree sound, a new 3D-knit textile mesh cover, and integration with Google TV Streamer. It also comes with Gemini for Home, Google's next-gen voice assistant, which offers advanced features like Gemini Live and multistep commands. However, some features require a subscription to Google Home Premium.

A data leak known as FortiBleed has exposed Fortinet and FortiGate VPN credentials for 73,932 devices worldwide. The leak includes usernames, email addresses, and plaintext passwords, and affects organizations across nearly every major industry sector.

A study found 41% of cybersecurity leaders cite AI-powered attacks as a top challenge, with alert fatigue and validating risks also being major concerns. The survey highlights the need for effective threat intelligence and continuous threat exposure management to tackle AI-driven threats.

Anthropic, an AI startup, has joined the Frontier carbon removal coalition, contributing to a new $915 million tranche of funding. This marks the first time an AI startup has joined the group, which aims to remove carbon from the atmosphere. The new funding brings the total to $1.8 billion and will help support various carbon removal technologies.

Cybercriminals have compromised tens of thousands of Fortinet firewalls and VPNs used by major companies worldwide. The hacking campaign, dubbed FortiBleed, uses automated tools to scan for exposed devices and breaks in using previously known passwords. Over 73,000 unique Fortinet URLs have been hacked, affecting companies like Accenture, Comcast, and Oracle.

The US government has banned Anthropic's most advanced AI models, Fable 5 and Mythos 5, citing national security concerns. However, experts argue that the ban is misguided and the legal basis for it is shaky. The ban has sparked controversy and raised questions about the government's motives and the impact on the AI industry.