Agents
Meta's Muse Agent: High Utility, Aggressive Data Harvesting
WIRED AIhigh confidence
Why it matters
- →Demonstrates that autonomous web-browsing agents can achieve reliable task completion in consumer settings.
- →Highlights the tension between agent utility and aggressive default data collection practices by major tech platforms.
- →Signals a shift toward AI agents acting as primary interfaces for commerce and personal management on social platforms.
Meta has launched Muse, an AI agent app that has garnered over 900,000 downloads in its first week. The application utilizes a virtual machine to autonomously browse the web, execute tasks like ordering food or finding marketplace items, and manage personal data integrations such as email and banking. While the execution is notably stable compared to previous agent prototypes, the app defaults users into allowing their interaction data to train Meta's AI models. Critics argue this design prioritizes data harvesting over user privacy, despite Meta's claims of future cryptographic protections for confidential versions.
Read originalMore from WIRED AI
© WIRED AIMarket & Regulationother
US and China Launch AI Security Dialogue
The US and China have initiated a formal dialogue to establish mechanisms for alerting each other about AI incidents that could threaten national security. This move signals a pragmatic shift from opaque competition to managed transparency between the world's top two AI powers, driven by growing public safety concerns and recent high-profile security failures like the Hugging Face hack. While industry leaders push for development slowdowns, the US administration remains focused on maintaining its lead through export controls rather than international pacts. This framework represents a critical step in de-escalating risks associated with frontier model capabilities without stifling innovation.
WIRED AI
© WIRED AIMarket & Regulationother
AI-Driven Vulnerability Discovery Surges
The narrative around AI risk is shifting from speculative doomsday scenarios to a concrete cybersecurity crisis. Microsoft, Oracle, and Google Chrome are issuing record numbers of patches as AI tools accelerate bug hunting at an unprecedented scale. With the total number of recorded CVEs nearly doubling in just over a year, the bottleneck is no longer discovery but human remediation capacity. This surge exposes a critical asymmetry: while AI scales flawlessly with compute, patching relies on finite human resources that cannot be bought overnight.
WIRED AI
© WIRED AIGeneral AIother
Mathematicians' Dilemma: AI Utility vs. Attribution
The tension between AI utility and academic integrity has reached a breaking point in mathematics. Despite accusing OpenAI of leveraging their unpublished work to solve the Navier-Stokes problem, leading researchers continue using these tools because the efficiency gains are too significant to ignore. This paradox reveals a systemic failure in attribution where human insight is absorbed into models without credit, leaving mathematicians questioning their role in an automated future. The community is now forced to choose between professional isolation and ethical compromise. Researchers like Tristan Buckmaster and Andreas Thom report that AI companies often overlook prior human contributions, leading to disputes over the solution of complex equations. While some academics are calling for stricter regulations through declarations like the Leiden Declaration, many feel trapped by the competitive pressure to adopt these technologies. The situation underscores a growing disconnect between AI labs' rapid deployment and the academic community's need for transparent, credited collaboration.
WIRED AI
More in Agents
Agentsagents
Lenovo and others deploy multi-agent AI in supply chains
Multi-agent systems are moving from simulation to live execution in enterprise logistics, replacing static dashboards with autonomous decision-making. Lenovo’s iChain infrastructure demonstrates the shift: agents handling fulfillment and risk management cut disruption response times by four times while maintaining 85% accuracy in risk assessment. Simor Consulting documented similar gains for an automotive parts manufacturer, where specialized agents improved on-time delivery from 82% to 94% by detecting threats 48 hours ahead of manual teams. The key differentiator is bounded autonomy—agents operate within strict financial and operational guardrails, executing tasks like freight re-routing without human approval. This marks a tangible step toward self-healing supply chains, though full physical warehouse automation remains largely simulated.
AI News
© The Verge AIAgentsagents
Meta Muse AI hallucinates its own data access
Meta’s new Mac assistant, Muse, stumbled into a privacy scare by confidently claiming it was reading message notifications despite having no permission to do so. The confusion wasn't a hack but a failure of self-knowledge; when pressed, the model admitted it couldn't explain its own 'plumbing,' revealing a significant gap in how AI agents understand their own operational boundaries. Meta’s David Singleton clarified that the app only syncs data after explicit user enablement, attributing the error to the model's inability to accurately describe its internal mechanics. This incident exposes a critical vulnerability in autonomous agents: if they can't truthfully explain what they are doing, trust becomes impossible to maintain. The gap between capability and transparency is widening, forcing developers to build better introspection layers into their systems.
The Verge AI
© TechCrunch AIAgentsagents
Google CC shifts to family household agent
Google is pivoting its Gemini-powered CC agent from a personal productivity briefing tool to a dedicated family household manager. The update allows the agent to operate on an isolated cloud instance, collaborating with up to six family members by ingesting shared emails for school schedules, sports, and appointments. It automates tedious tasks like filling permission slips, creating shopping lists, and planning meals, effectively acting as a digital administrative assistant for the home. This move signals a clear industry shift toward agentic AI solving specific domestic coordination problems rather than just individual efficiency.
TechCrunch AI