Topics
Latest
AI
Amazon
Image Credits:Jakub Porzycki/NurPhoto / Getty Images
Apps
Biotech & Health
Climate
Cloud Computing
Commerce Department
Crypto
Enterprise
EVs
Fintech
fund-raise
Gadgets
Gaming
Government & Policy
Hardware
Layoffs
Media & Entertainment
Meta
Microsoft
privateness
Robotics
protection
Social
Space
startup
TikTok
Transportation
speculation
More from TechCrunch
Events
Startup Battlefield
StrictlyVC
newssheet
Podcasts
video
Partner Content
TechCrunch Brand Studio
Crunchboard
get through Us
Microsoft ’s close better half and collaborator , OpenAI , might be suggesting that DeepSeekstole its IP and violated its terms of service . But Microsoft still wants DeepSeek ’s shiny new models on its swarm political platform .
Microsoft on Wednesday announced thatR1 , DeepSeek ’s so - called abstract thought mannequin , is available onAzure AI Foundryservice , Microsoft ’s program that brings together a telephone number of AI services for enterprises under a single banner . In ablog situation , Microsoft said that the version of R1 on Azure AI Foundry has “ undergone rigorous crimson teaming and prophylactic evaluations , ” include “ automated assessment of model behaviour and extensive security critique to mitigate possible jeopardy . ”
In the near future , Microsoft said , customers will be able-bodied to use “ distil ” flavors of R1 to operate locally onCopilot+ personal computer , Microsoft ’s brand of Windows computer hardware that meet certain AI eagerness essential .
“ As we preserve amplify the model catalog in Azure AI Foundry , we ’re excited to see how developer and initiative leverage [ … ] R1 to undertake real - universe challenges and deliver transformative experiences , ” continued Microsoft in the Emily Post .
The increase of R1 to Microsoft ’s cloud services is a curious one , see that Microsoftreportedly initiated a probeinto DeepSeek ’s potential abuse of its and OpenAI ’s service . According to security research worker work for Microsoft , DeepSeek may have exfiltrated a turgid amount of information using OpenAI ’s API in the fall of 2024 . Microsoft , which also happen to be OpenAI ’s largest shareholder , advise OpenAI of the untrusting activity , per Bloomberg .
But R1 is the lecture of the town , and Microsoft may have been persuaded to bring it into its cloud sheepfold while it still make allure .
Unclear is whether Microsoft made any modifications to the simulation to meliorate its truth — and battle its censorship . According to a testby info - reliableness organisation NewsGuard , R1 provides inaccurate answers or non - answer 83 % of the time when ask about news - related topic . A separate testfound that R1 refuses to answer 85 % of prompts related to China , possibly a effect of thegovernment censoring to which AI models develop in the res publica are subject .
