Microsoft cracks down on employee AI use after one worker spent $28,000 in 28 days
THE SO WHAT
Unmetered internal AI access is turning into surprise opex—$28,000 in a month from one employee is a governance failure, not a usage win. This week, put hard budgets, rate limits, and approval flows on internal AI tools before finance does it for you.
READ THE SOURCE
MORE FROM THE WIRE
OpenAI issues call for 'collective action' on cyber defense: 'We have a limited window'
When model providers start talking about a “limited window” on AI‑enabled cyber risk, the bar for “reasonable security” just moved. If you’re shipping AI features, assume regulators and insurers will treat this kind of joint statement as the new standard of care and document what you’re doing about it now.
Applied AIAs AI reshapes business, the fractional CTO finds its moment
The gap between agentic AI ambition and governance maturity—21% mature vs 74% planning to use agents by 2027—is creating a market for part‑time senior technical adults in the room. If you’re a non‑technical founder or board, a fractional CTO may be the fastest way to avoid sleepwalking into ungoverned agent deployments.
Applied AIOpenAI, Anthropic, Google, and 100 other companies call for action to defend against rogue AI
A hundred‑plus vendors co‑signing a “rogue AI” cyber memo is the industry telling boards: you’re on notice. Treat AI‑enabled attacks as a 2026 planning assumption—update threat models, vendor questionnaires, and incident playbooks accordingly.
Meta memo reveals what its new 'Hatch' AI agent can do, from booking a restaurant to finding a dog sitter
Meta pitching Hatch as an agent that “can do anything you can do online” is a clear move to own the personal‑task execution layer, not just chat. If your product lives in consumer workflows—bookings, errands, local services—assume those flows will be intermediated by agents and design for agent‑to‑API, not just user‑to‑UI.