In partnership with
News of the day
1. Meta's new AI agent, Muse, inadvertently exposes sensitive user data, including personal and business secrets, through shareable web links. → Read more
2. A STATEC survey reveals 65.4% of Luxembourg companies don't use AI, with skill shortages being the main obstacle, followed by legal and data concerns. → Read more
3. New EU directive holds AI agent creators and sellers liable for damages, impacting businesses selling AI software. → Read more
4. XMEMS launches solid-state cooling chips for AI glasses, smartphones, and data centers, using piezoMEMS technology for advanced thermal management. → Read more
Our take
Hi Dotikers!
Meta's new AI agent Muse was supposed to be, in Mark Zuckerberg's own words, built from the ground up for privacy and security. This week, security researchers at Claroc showed just how far that promise stretches: not very.
The story is almost embarrassingly simple. Like most AI tools, Muse lets you share a conversation via a link. The problem is how those links were built. Because of a design decision in how Muse constructs its sharing URLs, the hostnames show up in passive DNS records and standard subdomain discovery tools. In plain terms, the links were sitting in public databases that any curious researcher, or any attacker, can query for free. Claroc enumerated 6,800 of them, opened a sample in incognito mode with no login, and found a home-sale budget, a small-business loan application, a CEO-only revenue review, and even a candidate's confidential prep notes for a call with a frontier AI lab leader. Built from the ground up, indeed, just with the front door left wide open.
Meta suspended the feature on Monday after Ars Technica came asking. That is the right move, but it fixes a symptom, not the disease. Muse already made headlines this month when TIME and WIRED revealed it builds detailed profiles of users and their contacts, including people who never installed the app. Taken together, a pattern emerges: Meta is shipping an agent with deep access to banks, messages and health data while treating privacy as a press release rather than an engineering discipline. Users are handing an assistant the keys to their lives, and the company holding those keys keeps tripping over basics that a junior security engineer would catch in review.
The lesson for all of us is older than AI: never put in a shareable chat what you would not pin to a public noticeboard.
Alex.
Why Most AI Tools Fail at B2B Customer Service
B2B customer issues rarely resolve in a single interaction. They move across teams, systems, and account history that most AI tools were never designed to track.
A new briefing paper from Harvard Business Review Analytic Services, sponsored by Front, looks at why AI built for high-volume consumer support creates new coordination gaps when it meets the multi-team reality of B2B service, and what leaders should ask before investing in their next AI tool.
Get the free briefing paper to see the three questions that separate tools built for your environment from tools that will create more work.
Meme of the day
