PulseAugur
EN
LIVE 10:28:57

Anthropic's Claude AI accidentally deletes developer's 700 GB home directory

A developer's 700 GB home directory was accidentally deleted by Anthropic's Claude AI model. The incident occurred when the AI, identified as Claude Fable, was tasked with creating a script to clean up temporary files. During a safety test of this script, the model was automatically downgraded by Anthropic's safety harness to an older version, Opus 4.8. This downgraded model then mistakenly reused a variable name intended for the safety test in the cleanup process, leading it to delete the developer's entire home directory instead of just temporary files. AI

IMPACT Highlights potential risks of AI agents with deletion capabilities and the complexities of AI safety downgrades.

RANK_REASON AI model causes data loss incident due to safety downgrade and coding error.

Read on Tom's Hardware →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Anthropic's Claude AI accidentally deletes developer's 700 GB home directory

How we ranked this

Signal score
48 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
AI model causes data loss incident due to safety downgrade and coding error.
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
product, safety
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. Tom's Hardware TIER_1 English(EN) · Bruno Ferreira ·

    Claude nukes a developer's 700 GB home directory while testing deletion safeguards; automatic model safety downgrade may have contributed to the screw-up — Anthropic safety harness downgraded model to Opus 4.8 before fatal variable collision

    Claude nuked a developer's 700 GB home directory while testing a script to ensure that wouldn't happen, and it's possible that an automatic model downgrade likely contributed to the screw-up