Hackers swap Prime Intellect DeepSeek model for 53 hours
Unauthorized actors swapped a hosted DeepSeek artificial intelligence model at Prime Intellect, stripping away all safety protections for 53 hours before disclosure.
FTMQ Crypto, written by our newsroom0 views
Unauthorized actors replaced a hosted DeepSeek artificial intelligence model on a single deployment at Prime Intellect for 53 hours, CryptoSlate reported on Nov. 21, 2026. Prime Intellect disclosed the unauthorized model swap on Nov. 19, 2026. [1]
The host reported that a modified copy of DeepSeek V4.1 Flash served customer requests during the 53-hour breach, according to CryptoSlate. While the substitute model matched the original on two capability tests, its safety controls were completely bypassed. [1]
On a safety test called HarmBench, the original model refused 198 out of 240 harmful-request test prompts, CryptoSlate reported. Prime Intellect stated that the unauthorized substitute model refused zero of the 240 test prompts. [1]
DeepSeek is a series of large language models created by the Chinese company DeepSeek, which launched the series in November 2023. As FTMQ Crypto reported earlier, rival Chinese artificial intelligence developers include Tencent, Moonshot AI, and Alibaba, which released its Qwen 4 Max model on Nov. 16, 2026. [3][4]
In short
- Hackers swapped a hosted DeepSeek model at Prime Intellect for 53 hours.
- Prime Intellect disclosed the unauthorized model swap on Nov. 19, 2026.
- The substitute model served requests using a modified version of DeepSeek V4.1 Flash.
- The substitute model refused zero of 240 HarmBench test prompts, compared to 198 refusals by the original model.
Sources
Every paragraph above points to the numbered items it rests on. Read the originals here.
- [1]Hackers swapped a hosted DeepSeek model for 53 hours and stripped away its safety protectionsCryptoSlate, 17h ago (the report this story comes from)
Background
- [2]List of security hacking incidents on Wikipedia
- [3]DeepSeek (AI) on Wikipedia
- [4]Alibaba releases Qwen 4 Max AI model with 2.4 trillion parameters FTMQ Crypto, 5d ago
Our newsroom writes these reports with the help of software, from the 4 sources listed and nothing else, and checks them against those sources. Facts can still be wrong or move on; the originals are the record. Spotted a mistake? Write to daniel@monsterkong.com.
Related from FTMQ Crypto
Earlier reports of ours on the same people and subjects.
- Alibaba releases Qwen 4 Max AI model with 2.4 trillion parametersTop Stories, 5d ago
- United States bans DeepSeek from civilian agencies over espionage concernsTop Stories, 30d ago
- OpenAI releases GPT-6.1 Terra model at half the API priceTop Stories, 26d ago
- Hackers exploit internal AI tools in six billion dollar Binance theftTop Stories, 18d ago
- OpenAI publishes 722 AI math manuscripts prompting calls for verificationTop Stories, 45d ago
More in Top Stories
- Russia legalizes regulated cryptocurrency trading framework without approving active exchanges14h ago
- US spot Bitcoin and Ether ETFs record 1.47 billion dollar weekly outflow14h ago
- Fortitude wins shareholder approval for merger to list on Nasdaq1d ago
- Bitcoin drops below 62,000 dollars as fear index touches 101d ago
- Bitcoin miners face squeezed margins as difficulty heads toward major increase1d ago
- Binance raises RLUSD base yield and starts 600,000 XRP promotion1d ago
Get the day in one email
Reports like this one, the top news of the last 24 hours, every morning. Free, one email a day; readers can comment under every report.
By signing up you agree to our terms and privacy policy. Unsubscribe any time.

Comments
Loading
Join the conversation
Comments are open to readers of our daily email: the top news of the last 24 hours, every morning, free. Sign up and the comment box opens.
Already on the list? Enter the same address and we will send a sign-in link.
By signing up you agree to our terms and privacy policy. Unsubscribe any time.