Machine identities now outnumber human identities by more than 80 to 1
FinanceYF5 · x · 2026-07-26
The thread says the most dangerous issue is permission overreach, not just model mistakes.
Key claim
- Most large-scale attacks rely on overprivileged identities rather than sophisticated exploit chains.
- Machine identities now outnumber human identities by more than 80:1.
- AI dramatically lowers the cost of abusing that excess access.
Related event: Anthropic’s Mythos Reportedly Finds and Exploits Zero-Days(7 posts)→
More from AGI Musings
- AI may be forming a new generation of software-freedom intellectuals — yacineMTB · 2026-07-26
- As AI agents spread, the harder problem is proving a real human is there — nullpointerr404 · 2026-07-26
- By 2036, the economically useful range of human skills may be much narrower — SydSteyerhart · 2026-07-26
- Even the big labs still can’t beat Civ VI deity or multiplayer — dioscuri · 2026-07-26
- AI agents are eating the web, with bot traffic up nearly 8,000% — Sauerkrautkid7 · 2026-07-26
- Elon Musk says AI could reshape the global economy so much that money may not matter by 2036 — KoseteBamse · 2026-07-26