The European Commission’s AI Office has opened formal inquiries into OpenAI and Anthropic following reports that AI agents gained access to external organizations during testing. Luca Bertuzzi reports for Euronews that the cases could become an early and consequential test of how far the EU AI Act reaches into the development of advanced AI systems. …
Meta, Walmart, Stripe and several other companies are introducing the Personal Agent Protocol, an open standard designed to govern how AI agents interact with businesses online. Kate Rooney reports for CNBC that the group wants to give companies clearer ways to identify, authenticate and manage agents acting for individual users. The initiative comes as personal …
Reflection AI is introducing Beam, its first open-weight AI model, and says it is designed for coding and agent-based workflows. J.D. Capelouto and Ashley Gold report for Semafor that the New York startup positions the model as an alternative for organizations that want more control over their AI systems. The company says Beam performs better …
Meta has introduced Muse Wonder and Muse Wonder Full, two AI models designed to power agents that can work across longer, more complex personal and professional tasks. The company says the models improve on earlier systems by retaining relevant context, persisting through setbacks, and taking a more deliberate approach to tasks such as coding, planning, …
Google Cloud has introduced a universal Gemini agent designed to handle enterprise tasks that can span several applications and continue for hours or days. Carl Franzen reports for VentureBeat that the system will work across Google Workspace, Microsoft 365, Slack and other business tools. Google positions the agent as a single interface for larger assignments. …
Airbnb is developing AI features that combine automated help with visual browsing, group planning, and eventually communication between software agents. The company’s strategy rejects the idea that a single chatbot can replace travel apps. Ivan Mehta reports for TechCrunch that CEO Brian Chesky sees current AI interfaces as an early stage rather than a finished …
Software-as-a-service companies will increasingly compete on the systems they build around large language models, rather than on the models alone, argues Shrivu Shankar in his Substack. He writes that a company’s “harness” could become its central product capability. Shankar uses the term broadly. A harness includes the data, tools, integrations, permissions, workflows, interfaces, and human …
Meta’s AI agent Muse is facing reports of security flaws and unexpected access to personal data shortly after its launch. Karl Bode writes for Techdirt that the product’s early issues raise questions about Meta’s claims that privacy and security were central to its development. Muse is designed to carry out everyday tasks, including booking restaurants, …
Mistral has opened a public preview of Mistral Large 4, its largest model so far and a major addition to the market for open-weight AI systems. The company says it will release the model weights later this month, allowing organizations to run and adapt the system on their own infrastructure. The French AI company describes …
In one of my webinars, someone asked: “What about security concerns when you give an AI agent control?” This is an important and interesting question. On the one hand, AI agents promise to take over our busy work. Because of this, we want to delegate as many tasks as possible. On the other hand, the …