Uber burned through its entire 2026 AI budget in four months, so it capped every employee at $1,500 monthly per agentic coding tool.

Why it matters: It's the first major public signal from a big tech company of what enterprise AI coding actually costs at scale, and it points to a real governance gap: companies set AI budgets in 2025 before anyone knew how fast agents would burn tokens.

• At two tools per engineer and $36,000 annually per head, the cap runs to about 11% of the median Uber software engineer compensation package ($330,000, per Levels.fyi). That's a meaningful number, not a rounding error.
• The limit applies only to agentic coding tools like Cursor and Claude Code, not to chat interfaces, which tells you where the spend actually is.

Go deeper: https://simonwillison.net/2026/Jun/3/uber-caps-usage/#atom-blogmarks

AI Attackers Are Getting Harder to Spot, and MITRE ATT&CK Isn't Keeping Up

Anthropic studied 832 banned malicious accounts over a year and found that AI is shifting cyberattacks deeper into the kill chain, and the security frameworks defenders rely on can't see it.

Why it matters: The share of actors classified as medium risk or higher jumped from 33% to 56% in one year. Old signals like technique count and tool choice no longer reliably separate low-risk from high-risk actors, because AI now does the skilled work on their behalf.

• The clearest differentiator for high-risk actors is where they apply AI: post-compromise operations like lateral movement, account discovery, and privilege escalation, not initial access. But that signal is eroding as the broader population catches up.
• MITRE ATT&CK has no entry for agentic orchestration, the behavior where a model chains attack stages together with minimal human input. Anthropic says it's in talks with MITRE to close that gap.

Go deeper: https://www.anthropic.com/news/AI-enabled-cyber-threats-mitre-attack

Anthropic Formalizes Its Partner Tiers With New Services Track

Anthropic added a three-tier Services Track to the Claude Partner Network, giving enterprise buyers a published, verified scorecard for every consulting firm in the ecosystem.

Why it matters: With firms like Accenture (30,000 trained), Deloitte (470,000 people with access), and Cognizant (350,000 associates) now building practices around Claude, the ecosystem is big enough that "partner" needs to mean something. The tier structure is Anthropic's answer.

• Tiers are built on three measurable criteria: certified practitioners active in the last 90 days, customers running Claude in production, and public customer stories. Size doesn't move the bar; a ten-person shop and a global consultancy meet the same requirements.
• A new Claude Partner Hub publishes each firm's standing publicly and connects to Claude via MCP, so a partner can ask directly where they stand against the next tier.

Go deeper: https://www.anthropic.com/news/services-track-partner-hub