FTC Probing OpenAI and Anthropic Over Product Safety Concerns
THE SO WHAT
US regulators are now treating frontier AI safety as a live consumer protection issue, not a hypothetical—especially after security incidents. If you ship AI features, assume “product safety” will expand to include model behavior, data handling, and misuse pathways, and get your documentation and incident playbooks in order.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIAll the latest news on Meta’s cute, creepy Muse AI agent
Meta is turning Muse into a commerce and data firehose—email, browsing, and payments all routed through a single assistant. If your customer relationship lives in email or the browser, assume you’re now competing for surface area with an agent that can execute purchases directly.
Applied AIGoogle reportedly tests paying publishers for AI search results
Google piloting payments to ~100 publishers for AI search snippets is an attempt to price in content rights before regulators or courts do it for them. If you rely on search traffic, watch the contract terms closely—this is the template that may define how much upside you see from AI overviews versus traffic cannibalization.
Applied AIElevenLabs opens Brussels office and plans to triple its Belgian team
ElevenLabs planting a flag in Brussels and tripling headcount near EU institutions is as much a regulatory move as a commercial one. If you build AI infra or primitives and sell into Europe, assume proximity to policymakers and local teams is becoming part of the competitive stack, not a nice-to-have.
Applied AIAmex's New AI Agent for Expenses
Expense management is becoming an always-on agent layer, not a quarterly clean-up ritual—Amex wiring agentic tools directly into corporate cards is a distribution play as much as an efficiency story. If you sell into finance or T&E, assume your workflows will be competing with embedded card-native agents inside 6–12 months.