Topics
Latest
AI
Amazon
Image Credits:David Tramontan/SOPA Images/LightRocket / Getty Images
Apps
Biotech & Health
Climate
Image Credits:David Tramontan/SOPA Images/LightRocket / Getty Images
Cloud Computing
Commerce Department
Crypto
Enterprise
EVs
Fintech
Fundraising
gizmo
punt
Government & Policy
Hardware
layoff
Media & Entertainment
Meta
Microsoft
Privacy
Robotics
certificate
Social
Space
Startups
TikTok
transferral
speculation
More from TechCrunch
outcome
Startup Battlefield
StrictlyVC
newssheet
Podcasts
Videos
Partner Content
TechCrunch Brand Studio
Crunchboard
get through Us
AI chatbot DeepSeek has take the humans by storm , top out app store chartsand sending Silicon Valley and Wall Streetinto nuclear meltdown .
An offshoot of Chinese hedge stock High - Flyer , DeepSeek ’s cost - to - execution ratiomakes for impressive readingcompared to incumbents such as OpenAI . However , reportshave increasingly documentedsome of the thing the AI chatbot is unwilling to hash out , including thorny subject like theTiananmen Square protest and mass murder and Taiwan .
Anew reportfrom the folks atPromptFoo , an Andreessen Horowitz - backedstartup that helps companies find vulnerabilities in AI program , has get that DeepSeek ’s R1 abstract thought model refused to suffice some 85 % of 1,360 sensible - topic “ prompts , ” and the model ’s canned responses were replete with an “ over - the - top nationalist tone ” often offered or else .
The researchers also noted that DeepSeek can be easy jailbroken , suggesting that the Chinese laboratory has “ go through CCP [ Chinese Communist Party ] censoring in a crude , blunt - force way . ”
The full dataset of sensitive prompts areavailable on Hugging Face .
