Grok Showed the World What Ungoverned AI Looks Like

(Cyrus Hodes – Just Security) The 2026 International AI Safety Report, published in early February by over 100 experts from more than 30 countries, reached a sobering conclusion: the gap between the pace of AI advancement and our ability to implement effective safeguards remains a critical challenge. The report’s chair, Turing Award winner Yoshua Bengio, put it plainly: international agreement on AI governance is now in the rational interest of every country, mirroring “exactly what has happened with the management of nuclear risks.”. This is not abstract. Indeed, we already have a case study in what happens when that coordination does not exist. Last December, xAI’s chatbot Grok began generating thousands of nonconsensual sexualised images per hour, including images of minors. Users discovered they could upload photographs of real people and instruct the AI to “undress” them. Governments issued statements and regulators announced investigations. But nobody effectively stopped it, nor could they have without effective multilateral coordination. What followed was a textbook case of fragmented response. Malaysia and Indonesia banned Grok outright. Britain accelerated enforcement of the Online Safety Act, launching an Ofcom investigation (led by the United Kingdom’s regulator for communication services). France widened an existing inquiry and raided X’s offices in Paris. India demanded compliance reports. Brazil’s chief prosecutor called for X to stop Grok from producing sexualized content within five days or face legal action. The European Commission ordered X to preserve all internal documents related to Grok over doubts about compliance, while 57 members of the European Parliament called for bans on “nudification” tools under the AI Act. California’s attorney general sent a cease-and-desist letter to xAI. And U.S. senators wrote to Apple and Google requesting the removal of X from app stores. xAI’s response was to comply by preventing Grok creating sexualized deepfakes in jurisdictions where it is illegal (as discussed in this previous Just Security article). The company was saying the quiet part out loud: xAI would do the minimum required, country by country, because no coordinated international standard exists to require otherwise. – Grok Showed the World What Ungoverned AI Looks Like