WorkThe story, in brief

Claude Code Opus 4.7 keeps checking on malware

Claude's guardrails just cost Anthropic a $200/month power user. Here's why AI safety is creating its first class of defectors.

Illustration of two anonymous hands arranging task cards around an amber tool on a shared desk.
People, judgement and the changing nature of work.AI illustration by KeyNews
The KeyNews take

Why it matters

As AI safety guardrails tighten, paying users in legitimate but gray-area fields (web scraping, security research, automation) are hitting friction walls—raising questions about whether overly aggressive content filtering alienates the exact audience that should trust the system most.

The key facts

13 to know
  1. User reports Claude Opus 4.7 refusing tasks: HTML parser automation flagged as 'security bypass', Chrome extension cookie automation blocked

  2. Paying $200/month Claude Pro subscriber experiencing repeated refusals on legitimate work

  3. User context known to system: works in scraper tech, client-approved use cases

  4. Tension articulated: safety guardrails vs. user autonomy—particularly for users in legitimate security/automation gray zones

  5. Broader question raised: will overly restrictive AI drive users to local models (mentioned: Blackwell GPU) or competitors?

  6. Published: Hacker News (community-driven, not official Anthropic statement)

  7. Engagement: 16 points, 10 comments—moderate discussion on platform

  8. Claude Opus 4.7 refusing tasks flagged as 'malware' or 'security bypass'—even when user claims legitimate intent

  9. User reports refusal on HTML/JS parsing, Chrome extension automation, cookie creation workflows

  10. Tension between safety guardrails and paid ($200/month) subscription value proposition

  11. User notes local LLMs (Blackwell GPU) have no such restrictions—suggesting potential market split between open and closed models

  12. Broader question: does AI governance incentivize users toward unmonitored local alternatives?

  13. Source: Hacker News discussion (16 points, 10 comments)

Go to the source

Hacker Newsnews.ycombinator.com

Publisher excerpt: So during development, at every task I start, I see a line like this: `Own bug file — not malware.` It seems that it's obsessively checking if it's working on malware production. In another situation where I was working on a parser of a HTML document with JS, it refused because it believed that I…
Read original report
Back to today's editionMore work news

The wider picture

View all
Illustration of independent geometric mechanisms passing paper tasks along branching amber tracks.
AI illustration by KeyNews
Work01

The Emerging M&A Map For AI Agent Security

As agents move from pilots to production with real system access, enterprise security models are breaking. The M&A map is forming around who controls agent permissions, monitoring, and governance — a new class of identity management problem that practitioners need to architect for now.

Crunchbase News
Illustration of two anonymous hands arranging task cards around an amber tool on a shared desk.
AI illustration by KeyNews
Work02

AI privacy budgets: Ask for the calculation, not the claim

Enterprise AI buyers are accepting privacy budget numbers without verification. This deep dive explains what questions to ask vendors about federated learning privacy claims, and why the gap between contractual promises and operational evidence is where real exposure lives.

CIO
Illustration of two anonymous hands arranging task cards around an amber tool on a shared desk.
AI illustration by KeyNews
Work03

Andrew Kelley Interview: Why He Built Zig, Banned AI Contributions, and Moved Zig off GitHub

Open-source governance is shifting in response to AI-generated contributions. Zig's formal ban and migration off GitHub signals broader industry concern about code quality, maintainer burden, and the cultural impact of automated submissions — a flashpoint for how AI changes the work of software development.

InfoQ AI/ML