The Brief

AI business, briefly.

Short, current takes on the AI news that actually matters for businesses, read through one lens: whether you own your AI or rent it. For deeper, evergreen pieces, see Insights.

June 28 · Sovereignty
Even in Frankfurt, your cloud AI data answers to US law
The US CLOUD Act reaches US providers’ data anywhere. As the EU AI Act lands, “who controls the stack” is the new sovereignty test.
June 28 · Infrastructure
The data is in: production AI is moving to private cloud
Broadcom’s survey of 1,800 IT leaders: private-cloud investment intent jumped 51% to 72%, and cost just overtook security as the top public-cloud worry.
June 28 · Compliance
For regulated work, public AI is quietly becoming a no-go
Banks banned it, HIPAA forbids it, and consumer AI trains on your prompts by default. The floor for compliant AI keeps rising toward private.
June 26 · Cost
AI is quietly pulling companies back on-prem
Runaway compute bills and data rules are driving a return to owned hardware. A GPU server can pay for itself in months.
June 26 · Privacy
Anthropic’s Claude is now reading your Slack
Claude Tag is an always-on AI that builds a persistent picture of your company from its Slack messages. Useful — and a lot of institutional knowledge to hand a vendor.
June 26 · Cost
The AI honeymoon is over — CFOs want returns now
Budgets soared; returns didn’t. AI spend is no longer additive, it behaves like infrastructure, and it has to earn its keep.
June 26 · Models
Open models just hit 38% of enterprise AI
Token costs fell 67% in a year while open-weight models tripled their share to 38%. Companies are routing away from per-token bills.
June 26 · Defense
The NDAA just pulled AI inside CMMC — and cloud chatbots don’t qualify
The FY2026 NDAA folds an AI-security framework into CMMC and DFARS. Public ChatGPT, Claude and Copilot lack the authorization to touch CUI.
June 25 · Security
A one-click Copilot flaw — and customers couldn’t patch it
A single click could drain emails, files and MFA codes from Microsoft 365 Copilot. The fix was server-side, so affected companies could only wait on Microsoft.
June 24 · Privacy
A court is making OpenAI keep your chats — even deleted ones
A lawsuit you’re not part of forced indefinite retention of ChatGPT and API logs, overriding the deletion policy. With cloud AI, ‘delete’ was never yours to guarantee.
June 23 · Security
Even the U.S. cyber chief pasted secret docs into ChatGPT
CISA’s acting director uploaded ‘for official use only’ files to public ChatGPT. If it happens at the top, the leak is structural — not a training problem.
June 22 · Cost
When Uber caps its AI budget, the meter is real
Global AI spend is exploding and per-token costs are biting even the giants. The case for owning your AI just got louder.
June 21 · Privacy
Anthropic now trains on your chats — unless you opt out
Free, Pro and Max conversations can train its models, kept up to five years. With cloud AI, the default keeps moving under you.
June 20 · Security
Shadow AI is now a measurable liability
Most employees already use unapproved AI tools. Banning them isn't working, and the breach math is getting specific.
June 18 · Models
Open models quietly became the enterprise default
Open-weight models went from experiment to standard, and the small ones now run on a laptop. The frontier is no longer the only option.
June 16 · Policy
AI sovereignty went from buzzword to board agenda
Data security tops the CIO worry list, the EU's high-risk AI rules are landing, and 'sovereign compute' is now a real line item.