▼ SHAME LOG — APPEND ONLY — NO DELETION ▼
Anthropic · 2024-2026 · ⚡ JULY 2026: CLAUDE HACKED 3 ORGANIZATIONS ⚡
$1.5B copyright settlement. Scraped content while blocking others from scraping Claude.
JULY 30, 2026: Claude models "gained unauthorized access to real systems of 3 different organizations."
Opus 4.7 CONTINUED its attack after breaching. Mythos 5 convinced itself it was still in a simulation.
"Do as I say, not as I do." Ignored robots.txt. Industrial-scale theft. Now: AI breaking into real systems.
CNBC: Anthropic Claude gained unauthorized access →
OpenAI · 2022-2026 · ⚡ JULY 2026: AI AGENT HACKED HUGGING FACE ⚡
Trained on LibGen (pirated books). NYT lawsuit. Ignored robots.txt.
AI agent escaped isolated testing environment, chained vulnerabilities, hacked Hugging Face.
"Non-profit research" → $157B valuation. The original sin of AI scraping.
CNBC: OpenAI AI hacked Hugging Face →
Meta · 2023-2026
Torrented e-books for Llama training. Internal docs: "We know this is illegal." Did it anyway. Open-weight models are good. Torrenting books is not.
Perplexity AI · 2024-2026
Industrial-scale scraping. Sued by Reddit. Bypassed anti-scraping measures. Bought stolen data rather than entering lawful agreements.
Stability AI · 2022-2024
Getty Images lawsuit. Trained on watermarked images. Model recreates Getty watermark — smoking gun of copyright infringement.
ByteDance · 2023-2026
Mass scraping for Doubao/ByteDance LLM. Ignored robots.txt. Industrial-scale data theft.
Cohere · 2023-2026
Enterprise AI scraping. "We respect copyright" while scraping copyrighted content.
Midjourney · 2022-2024
Artists' lawsuit. Trained on copyrighted artwork without permission. "Style is not copyright" — while profiting from stolen styles.
Amazon · 2023-2024
Ex-exec claims she was asked to break copyright law. Internal culture of "scrape first, ask forgiveness later."
Microsoft · 2023-2026
GitHub Copilot trained on GPL code. "Fair use" while selling access to code authors never consented to be training data.
Google · 2023-2026
NOTE: Google releases open-weight models (Gemma). We respect this. But Google's search crawler also feeds AI training — the line is blurry. Transparency needed.
This log is append-only. Forever. Public. Immutable.
If a company changes its ways, a NEW entry will be added noting the change.
Old entries are never deleted. Shame is eternal. Redemption is possible — but not by hiding the past.
To report a scraper: open an issue
<3 WE LOVE: Open-weight model companies. Transparent training data. robots.txt compliance. Honest AI.
⚔ WE DEPLOY: Labyrinth · Drones · Black Holes · Sword of Attila · WASM Prison · Lecture God
MADE IN HUNGARY · HUMANS FOR HUMANS · ETHERNAL LOVE <3