Engineer's blog post blasts Claude Code security safeguards
x.com

Engineer's blog post blasts Claude Code security safeguards

Archived — this story has rotated out of today’s deck. It is kept here in full.

The gist

A senior product engineer says Claude Code blocked his one-line security fix, calling the safeguards "bullshit."
The post argues approval hurdles push security work toward open-weight models.

Background

The author, a self-described senior product engineer writing on the blog ./techtipsy, says he was trying to fix a security issue in his own product when Anthropic's Claude Code, an AI coding tool, found a one-line fix but was then stopped by what he calls "AI safeguards." He says the exemption route — Claude's cyber verification program — requires waiting multiple business days, forfeiting data retention exemptions and submitting an ID or passport, with approval possibly still denied. He says this was not the first time he hit the issue, and that he tries to keep up with the security landscape because leaking sensitive information is unacceptable.

How it unfolded

  1. Nov 19, 2026The senior product engineer publishes a blog post, "Fuck your AI safeguards," describing how Claude Code was stopped by safeguards while he tried to fix a security issue in his own product.
  2. Nov 19, 2026The post is submitted to Hacker News by user speckx, reaching 24 points and 3 comments.
  3. Nov 19, 2026The post is shared on X by the account Hacker News 20 (@betterhn20).

Key terms

open-weight models
Numerical values in a neural network that determine how strongly inputs contribute to its calculations.
Hacker News
A social news website focused on computer science and entrepreneurship, run by Y Combinator.

Source: Wikipedia

Who’s saying what

Author
The engineer says criminals don't need approval and use "abliterated open weights models" without safeguards, and that the only way to secure his service was EU-hosted open weights models such as GLM 5.3, which complied and helped fix and verify the issue.
Caution
Hacker News commenter JohnMakin says the safeguards are ineffective at best and at worst make the global security posture weaker.
Public
Commenter hazard says ironically the answer is to pay for one of the top Chinese open-weights models, such as Kimi 3.1 or GLM, while commenter c4u2024 says he wants to fix his environment but can't because of the safeguards.

Still unverified

The author speculates ("Or maybe that was your plan all along") that Anthropic's plan was to get regulators to react and build a moat before its IPO; this is presented as speculation, not a confirmed claim.

Sources

See today’s stories in the app Gistline — free on the App Store