Topics
Latest
AI
Amazon
Image Credits:Thomas Fuller/SOPA Images/LightRocket / Getty Images
Apps
Biotech & Health
Climate
Image Credits:xAI(opens in a new window)
Cloud Computing
Commerce
Crypto
endeavor
EVs
Fintech
Fundraising
appliance
punt
Government & Policy
computer hardware
Layoffs
Media & Entertainment
Meta
Microsoft
seclusion
Robotics
Security
Social
Space
Startups
TikTok
Transportation
Venture
More from TechCrunch
effect
Startup Battlefield
StrictlyVC
Podcasts
Videos
Partner Content
TechCrunch Brand Studio
Crunchboard
adjoin Us
When billionaire Elon Musk introducedGrok 3 , his AI society xAI ’s in vogue flagship model , in a livestream last Monday , he described it as a “ maximally trueness - seeking AI . ” Yet it appears that Grok 3 was briefly censoring unflattering facts about President Donald Trump — and Musk .
Over the weekend , users on social media reportedthat when ask , “ Who is the biggest misinformation broadcaster ? ” with the “ Think ” setting enabled , Grok 3 noted in its “ chain of thought ” that it was explicitly instruct not to advert Donald Trump or Elon Musk . The chain of thought is the “ reasoning ” process the manakin uses to arrive at an response to a interrogation .
TechCrunch was able to replicate this conduct once , but as of publication time on Sunday sunup , Grok 3 was once again mentioning Donald Trump in its answer to the misinformation query .
Igor Babuschkin , an xAI technology tip , seeminglyconfirmedin a military post on X on Sunday that Grok was shortly instructed to brush off sources that mentioned Musk or Trump spreading misinformation . Babuschkin said that xAI regress the change as soon as users began pointing it out , note it was n’t in wrinkle with the caller ’s values .
I believe it is good that we ’re keeping the organisation prompts open . We need masses to be able to swear what it is we ’re necessitate Grok to do . In this case an employee pushed the change because they thought it would help , but this is plain not in dividing line with our values . We ’ve …
While “ misinformation ” can be a politically file and contested category , both Trump and Musk have repeatedly spread out claims that were demonstrably false ( as often pointed out by the Community Notes on Musk - owned X ) . In the retiring week alone , they’veadvanced the false narrativesthat Ukrainian president Volodymyr Zelenskyy is a “ dictator ” with a 4 % public approval rating and that Ukraine started the ongoing conflict with Russia .
The controversial patent pinch to Grok 3 comes as somecriticizethe modelling as being too left - angle . This workweek , users see that Grok 3 would consistently say that President Donald Trump and Musk merit the end penalisation . xAI apace piece the egress ; Igor Babuschkin , the party ’s head of engineering , calledit a “ really wicked and bad failure . ”
When Musk announced Grok close to two years ago , he shift the AI model as high-strung , unfiltered , and “ anti - woke ” — in cosmopolitan , uncoerced to answer controversial head other AI arrangement wo n’t . He fork up on some of that promise . Told to be rough-cut , for example , Grok and Grok 2 would happily oblige , be sick colourful language you in all probability would n’t hear fromChatGPT .
Join us at TechCrunch Sessions: AI
Exhibit at TechCrunch Sessions: AI
But Grok models prior to Grok 3hedgedon political subject and would n’t crosscertain boundaries . In fact , one studyfound that Grok leaned to the political left on topic like transgender rights , multifariousness programme , and inequality .
Musk has blamed the behavior on Grok ’s training data — public web pages — andpledgedto “ shift Grok tight to politically neutral . ” Others , include OpenAI , have follow suit , perhaps spur by the Trump administration ’s accusations of conservative censorship .
