Stay informed with free updates
Simply sign up to the Artificial intelligence myFT Digest — delivered directly to your inbox.
Elon Musk’s xAI is pushing its artificial intelligence model towards giving users more risqué answers than its risk-averse competitors, in an effort to attract users with adult content.
The start-up released its latest model, Grok 3, this month boasting comparable performance to competitors from OpenAI, Google and Anthropic. xAI’s model underpins its Grok chatbot, which has fewer “guardrails” than rivals, in an effort to align with Musk’s mission of being a “maximally truth-seeking AI”.
Over the weekend, xAI launched the ability to converse over voice with the chatbot, including preset personalities such as “romantic”, “sexy” and “unhinged” — the latter two labelled as “18+”. By contrast, OpenAI’s voice mode, which was released last year, offers a range of different voices and personalities without specific adult-centred experiences.
Musk’s move aims to drive engagement by offering a unique selling point over its competitors, according to people close to the company. They said that towards the end of training Grok 3, there was a pivot to allowing adult content and explicit responses.
“I have no doubt that Musk was seeing dollar signs and wanted to differentiate from OpenAI’s offering,” said one of the people.
They added: “It was meant to be a helpful assistant, maybe even a therapist, but . . . this is not going to encourage human relationships.”
Musk’s xAI has joined a global battle to release large language models that can be adopted by consumers and businesses. Grok is available as a standalone app as well as for users of X, the social media platform Musk bought for $44bn in 2022.
The Grok app has been downloaded about 4mn times since its launch, according to estimates from market intelligence group Sensor Tower. Its analysis shows that the week the latest model launched, daily active usage was up 330 per cent in the US.
Data from X is used to train the Grok model, while more sophisticated features are offered to paying premium subscribers of X.
According to two investors, the crossover between Musk’s X and xAI is part of the appeal to those who have backed the companies, suggesting the introduction of the chatbot has boosted engagement on the social media site.
X, formerly Twitter, has long allowed nudity and consensual pornography on the platform, in a divergence from rivals such as Facebook parent Meta. About 13 per cent of content on Twitter was not suitable for work, according to a Reuters report from 2022.
Shortly after Musk took over the platform, he explored introducing a feature to allow users to offer videos of adult content behind a paywall similar to those posted on subscription site OnlyFans, according to people familiar with the matter. However, the move was halted over fears about hosting child exploitation on the platform.
Grok’s recent changes have also raised concerns over child safety. The app has a 12+ age rating on app stores, and the chatbot is available through X. However, multiple people close to the company have raised concerns that users can access the 18+ presets without checks on age.
xAI’s terms of service state the product is “not directed at children or minors under the age of 13” but that “Grok could produce output that is not appropriate for all ages”. It added that parents of teenagers between 13 and 17 must agree to the terms and monitor usage. “The service may have content such as some suggestive dialogue, coarse language, crude humour, sexual situations, or violence,” it added.
Sexual or romantic encounters are becoming an increasingly common use case for generative AI chatbots, throwing up legal and ethical issues. Research from the University of Sydney business school found that half of respondents used AI for friendship, a third for sex or romance and nearly 20 per cent for counselling.
“AI companions offer a powerful business advantage by fostering deep emotional connections that drive high user engagement, loyalty and repeat visits,” said Raffaele Ciriello, a lecturer in business information systems at the University of Sydney. “This emotional bond boosts customer retention while enabling companies to collect sensitive personal data on user preferences.”
He added that AI companions could be “highly addictive” and lead to unhealthy dependence, calling for “cautious implementation and robust regulation, especially as AI chatbots like Grok 3 increasingly cater to human companionship and intimacy”.
Grok 3 also came under fire this week when it emerged the system had been instructed to “ignore all sources that mention Elon Musk/Donald Trump spread misinformation”.
Igor Babuschkin, xAI’s co-founder, said in a post on X that the change was made by “an ex-OpenAI employee that hasn’t fully absorbed xAI’s culture yet” and had been corrected.
Source link
