2026-05-15 18:47 UTC
Replies (5)
-
@1password@1password.social 2026-05-15 18:48
Our engineers used AI-assisted tooling the same way many teams use modern development tools: to help review, organize, and improve internal code. Every change was reviewed by engineers and went through our standard security and quality processes before being accepted. The security model behind 1Password remains the same. We cannot see the contents of your vault, and neither can anyone else. That has not changed.
-
@daniel@networked.systems 2026-05-15 18:53
@1password@1password.social Weird how „multi-million-line Go monolith“ turned into „internal systems“
-
@lumarel@ohai.social 2026-05-15 19:18
@1password@1password.social If you say it like that... I hope you increase the code reviews 10x if not even 100x, if the code interfacing with the vaults in any way starts to get produced with AI assistance, coding support for speed increase has no value compared to knowing the platform is and stays secure. This also makes me ask for a self-hosted option again, to limit the attack surface myself... (regional filtering, vpn only access, ...)
-
@virtualmarc@edgie.me 2026-05-15 20:14
@1password@1password.social Internal systems? In your blog post you explicitly write about a big monolith and production systems. You also state that you shift the engineering workflow from writing code to defining systems. With other words: Vibe coding. It doesn't matter that the AI has no access to customer data/our secrets/vaults/etc. If the software that handles this data (the backend infrastructure, the clients, browser plugins, cli, etc.) is written by AI, this is the incident waiting to happen. Be it a big security vulnerability (since AI is known to repeat common vulnerabilities in a study with Claude Code changes on public GitHub repositories) that exposes private customer data to third parties, or a complete data loss that happens all of the time to companies trusting AI agents. And at the speed AI generates code, and with that amount of code beeing generated, no human will ever (be able to) review all of that. This would make them way less productive and slower than writing the code by themselves and doing peer reviews. AI can analyse code, it can help find bugs or vulnerabilities, but it should never write code. Writing code is not what LLM's are good at.
-
@2Bfair@infosec.exchange 2026-05-16 10:15
@1password@1password.social "We hear your feedback clearly" Bullshit. You think we were born yesterday? That we've never worked at companies with marketing/pr/propaganda/executives?