Anthropic studied 832 banned accounts over a year and found AI is making less-skilled attackers far more dangerous, especially once they're already inside a network.

Why it matters: The old way of gauging attacker risk, counting techniques used and flagging which interface they logged in from, no longer works. AI lets low-skill actors run the same deep-network moves that used to require serious expertise.

• Medium-risk-or-higher actors jumped from 33% of the sample in the first six months to 56% in the second. The share doing post-compromise work (account discovery, lateral movement) is climbing while phishing use is falling.
• MITRE ATT&CK has no entry for agentic orchestration, where a model chains attack steps autonomously with minimal human input. That's exactly what a state-sponsored group did in a disrupted November 2025 operation. Anthropic is in talks with MITRE to update the framework.

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

Uber Caps AI Coding Tools at $1,500 Per Engineer Per Month

Uber blew its full 2026 AI budget in four months, so it's now capping each employee at $1,500/month per coding agent tool.

Why it matters: This is the first hard public number on what a large tech company thinks AI coding tools are actually worth, and it's a meaningful one: two tools at the cap equals $36,000/year per engineer, or about 11% of the median Uber software engineer's compensation package.

• The cap applies only to agentic coding tools like Cursor and Claude Code, not to other AI products. Each tool is budgeted separately.
• Simon Willison notes his own usage runs ~$1,000/month per provider, so most engineers would stay under the cap with room to spare. The real driver was the budget shock, not runaway individual use.

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

Anthropic Tiers Its 10,000-Consultant Partner Network

Anthropic is adding a three-tier structure to its Claude Partner Network, with public scorecards showing exactly how many certified practitioners and production deployments each firm has.

Why it matters: More than 40,000 firms applied to join since the March launch. The new Services Track gives enterprise buyers a way to tell real Claude practices from badge-collectors, with the same bar for a ten-person shop and a global consultancy.

• Tiers run from Select (10 certified individuals, 2 production customers) to Global Premier (1,000 certified individuals, 100 customers across three-plus regions). Promotions happen twice a year; demotions only at the annual December 31 review after 90 days' notice.
• Major SIs are already in: Accenture (30,000 professionals), Deloitte (470,000 people), and Cognizant (350,000 associates) are all building Claude practices.

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