In partnership with
Hey friends 👋
Today we've got Anthropic rolling out spending guardrails for enterprises freaking out about their AI bills, plus Illinois just passed one of the toughest AI safety laws in the country. Let's dig in.
Anthropic Just Admitted Your AI Spending Is Out of Control
Anthropic built its enterprise business on being the safe, careful AI lab. Now it's applying that same instinct to your budget.
Claude Enterprise rolled out three new features aimed squarely at CFOs and CIOs who've watched their AI inference costs spiral. Admins get notified when an org hits 75% and 90% of its usage limit, so they can raise the cap before things lock down. Employees get their own alerts at 75% and 95% of personal limits, with the ability to request an increase straight from the Claude app. Admins can now see cost alongside output for real ROI analysis, and they can set default models by role or org-wide, so people start on cheaper models and only upgrade when the task actually needs it.
This didn't come out of nowhere. Uber reportedly built internal leaderboards ranking employees by token usage and burned through its entire annual AI budget in four months. Databricks CEO Ali Ghodsi said last month that "people's hair is on fire" over inference costs. And Anthropic has more skin in this game than almost anyone. New Ramp data shows 77% of organizations adopting frontier LLMs are now using Anthropic models, up 40 points in a year, and Anthropic's revenue run rate jumped from $9 billion at the end of 2025 to $47 billion last month.
Here's why it matters. Anthropic makes more money the more tokens you burn, so building tools that help you burn fewer of them looks backwards on paper. But trust is the actual product here. If enterprises start associating Claude with runaway bills, that reputation damage costs Anthropic more long term than a few saved tokens cost it short term. This is a company betting its safe, responsible image is worth more than squeezing every last dollar out of the meter right now.
You've seen the AI demos. Viktor does it without you watching.
The AI tool you tried last quarter waited for a prompt, hallucinated a number, then asked if you'd like a summary.
Viktor opened a PR at 2am, rebased it against main, ran your test suite, and posted a note in #eng: "Two flaky tests in payments service, both pre-existing. Recommended merging after fixing them." Then drafted the customer reply for the support ticket the bug created.
That's 619K autonomous actions per day across 20,000+ teams. Not chat replies. Real work shipped to GitHub, Stripe, Linear, Notion, and 3,000+ other tools, from inside Slack and Microsoft Teams.
You don't supervise him any more than you supervise a senior engineer.
SOC 2 certified. Your data never trains models.
"It's what you probably originally thought AI was going to be when you first heard of it in sci-fi movies." Tyler, CEO.
Illinois Just Made It Harder for AI Companies to Regulate Themselves
While Congress sits on its hands, states keep stepping into the AI safety vacuum, and Illinois just went further than most.
Governor JB Pritzker signed SB 315 on Monday, a bill that passed unanimously and was backed by both OpenAI and Anthropic. It only applies to frontier AI developers pulling in $500 million or more in revenue, but for that narrow group, the rules have teeth. Companies have to publicly disclose their safety frameworks, submit to third-party evaluations, report critical safety incidents like model theft or user deception, and assess for catastrophic risks including loss of control, weapons development, or large-scale cyberattacks. Skip it and you're looking at a $1 million fine for a first violation and $3 million after that. The law takes effect January 1, 2027.
What makes this one different from the California and New York laws it builds on is the independent evaluation requirement. Self-reported safety claims are one thing. Having an outside party actually check your work is another. The timing is notable too, coming right after Anthropic's own rocky Mythos and Fable rollout and OpenAI holding back GPT-5.6, while multiple AI firms are reportedly in talks with the government about a voluntary standard for model releases.
Here's why it matters. Voluntary standards sound nice until a company decides not to volunteer. Illinois is a reminder that states aren't waiting around for AI labs to police themselves, and mandatory third-party audits are a meaningfully higher bar than anything on the table at the federal level right now. If more states follow this exact template, that becomes the real regulatory floor for frontier AI, whether Washington ever gets its act together or not.
A few more things worth knowing
Reddit says AI-powered detection now blocks 23 million spam views per day and catches 25,000 new spammy posts daily, and user exposure to spam dropped 20% from January to March
The same tools are being used to catch hate and violent content in English posts in under five seconds
New Oxford and Potsdam research found that LLMs used to write social posts on contested topics quietly shifted the meaning of those posts, even when explicitly told to preserve it, which could add up to real shifts in public opinion over time
That's what stood out to me today. Reply and tell me what caught your eye, I read everything.
Talk tomorrow,
Hatman 🎩