---
title: "OpenCode key breach; Anthropic sets up a public benefit fund"
date: 2026-10-18
updated: 2026-10-18T00:00:00Z
url: https://arun477.github.io/feed/2026-10-18/
author: Arun Arumugam
---

# OpenCode key breach; Anthropic sets up a public benefit fund

A quiet weekend for launches. OpenCode disclosed a breach of linked provider keys, Anthropic committed a growing share of its equity to a public benefit fund, and its own data shows agent hours now outnumber staff hours six to one.


- [OpenCode Console breach exposed some users' linked AI provider keys](https://x.com/opencode/status/2111519637049319818) (OpenCode, news, Security): Attackers tied to a large distillation campaign reached 1.4% of Console users who had OpenAI, Anthropic or Google keys linked, and some saw unauthorized usage. Affected users were emailed and refunds are underway; anyone who linked keys should rotate them.
- [Anthropic commits 5% of its shares to a new Public Benefit Fund](https://www.anthropic.com/news/public-benefit-fund) (Anthropic, news, Labs): Subject to stockholder approval, the stake can grow 2% for every 50% rise in value; with founder and staff pledges, about 20% of Anthropic's equity is now committed to public benefit. AI's labour-market impact is a named focus, and none of it goes to lobbying.
- [How employment is changing at Anthropic as agent hours pass human hours](https://www.anthropic.com/research/employment-at-anthropic) (Anthropic Research, paper, Research): Internal data shows agent hours outnumbering employee hours six to one, output per person rising after people adopt coding agents, and fewer junior technical hires. Human attention to review agent work is now the bottleneck.