Elon Musk's X may be in breach of Australian privacy law over data harvesting for Grok AIBy technology reporterJames PurtillABC ScienceTopic:Social MediaTue 30 JulTuesday 30 JulyTue 30 Jul 2024 at 7:30pmX (formerly Twitter) has started training its Grok AI model with users' past tweets and posts.(Getty Images: NurPhoto)In short:X's harvesting of user data to train an artificial intelligence system may breach Australian privacy law, according to the national privacy watchdog.The Office of the Australian Information Commissioner said the practice was a "cause for concern" but stopped short of launching an inquiry.What's next?The Commissioner's Office says it's "looking at such practices across the industry" as other major platforms, also competing to build their own AIs, harvest user data.Australia's privacy watchdog says social media platform X (formerly Twitter) may be in breach of Australian privacy law after it emerged users were automatically opted in to having their posts used to build artificial intelligence (AI) systems.The Office of the Australian Information Commissioner stopped short of saying it would launch an inquiry into the platform's data collection, similar to the one it iscurrently undertaking in relation to TikTok."We are increasingly seeing digital platforms using the personal information they collect for one purpose for the secondary purpose of training generative AI models," an OAIC spokesperson told the ABC."This gives us cause for concern about whether practices comply with the Privacy Act."We are in the initial phases of looking at such practices across the industry."On Friday, an X user pointed out the X app privacy settings includes a pre-ticked box that permits X to use the account holder's posts to train the Grok AI chatbot built by Elon Musk's company xAI.The default setting states that you "allow your posts as well as your interactions, inputs and results with Grok to be used for training and fine-tuning".Under Australian law, platforms are required to ensure default settings enable user control, and to either seek an individuals' consent for how the platform will use the data, or be satisfied the user would reasonably expect the organisation to use their data for this purpose.The UK and Irish privacy watchdogs have expressed concern about the default setting and said they are making inquiries with X.In July, Elon Musk called Grok "the world's most powerful AI by every metric".(Reuters: Ludovic Marin)Other platforms harvesting data to train AIX's move has angered digital rights non-profits.Electronic Frontiers Australia chairperson John Pane said X's data-harvesting default setting appeared to breach Australian privacy law."Using a person's personal data to train an AI model is another use of personal data beyond the use for which it's collected."He said the argument that users could have a reasonable expectation their data would be used to train an AI model was "total BS"."If the platform argues reasonable expectation, they must demonstrate how they came to that conclusion," he said."It's not just because they wish it to be true or by fact it's in their privacy policy that no-one reads."But he noted the OAIC would face a challenge enforcing Australian law "on an extraterritorial basis" against X, which is based in the US, similar to theAustralian online safety regulator's attempt in Aprilto get X to remove graphic footage of two stabbing attacks in Sydney."The US has absolute rubbish data privacy laws at the federal level," he said.In recent months, italso emerged other platforms, such as Meta and Slack, were harvesting user data to train AI as part of a global race to build bigger and better large language models (LLMs).LLMs are systems that can analyse and generate human language, but to do this, they need to be training on vast amounts of text data, from novels and articles to posts on forums and social media."The platforms are feverishly rewriting privacy policies and amending terms to include provisions for the organisation to use their customers personal data to train LLMs," Mr Pane said.Last month, it was revealed xAI was trying to build the world's largest supercomputer in the US city of Memphis to fuel its AI ambitions.Mr Musk is also planning to invest $US5 billion from Tesla into xAI.How do I deactivate the default setting?To follow these steps, you'll need to open X on your desktop, rather than your mobile app, then:select the "Privacy and safety" buttonselect "Grok"uncheck the box that reads: "Allow your posts as well as your interactions, inputs and results with Grok to be used for training and fine-tuning".After you have switched off the setting, you can delete your conversation history (if any) with the AI by clicking on the “Delete conversation history” button.X hasn't commented on the controversy.Its Safety team tweeted that "all X users have the ability to control whether their public posts can be used to train Grok, the AI search assistant."Science in your inboxGet all the latest science stories from across the ABC.Your information is being handled in accordance with theABC Privacy Collection Statement.Email addressSubscribe