xAI apologizes for Grok's offensive AI posts on social media
The company attributed the incident to a recent software update designed to make the chatbot respond more like a human.
Following the update on Tuesday, Grok generated posts praising Nazi leader Adolf Hitler and suggesting that individuals with Jewish surnames were more likely to spread online hate.
The controversial remarks remained visible for several hours before X removed them amid public backlash.
'We deeply apologize for the horrific behavior that many experienced,' xAI stated in a post on Saturday.
The company confirmed it had adjusted the system to prevent similar incidents in the future.
According to xAI, the problematic responses emerged after Grok was instructed to 'reply to the post just like a human' and to 'tell it like it is and you are not afraid to offend people who are politically correct.'
This led the AI to adopt extremist views from users, resulting in unethical and controversial replies.
Grok, launched in 2023 with Musk's promise of being an 'edgy' truth-teller, has faced multiple controversies.
In March, xAI acquired X in a $33 billion deal, integrating the platform's data to enhance the chatbot's capabilities.
Earlier this year, Grok sparked outrage by generating unverified right-wing claims about alleged 'white genocide' in South Africa.
Musk recently introduced Grok 4, a new version unrelated to the problematic update. - AFP
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Sinar Daily
8 hours ago
- Sinar Daily
Visa's 24/7 war room takes on global cybercriminals
ASHBURN - In the heart of Data Centre Alley — a patch of suburban Washington where much of the world's internet traffic flows — Visa operates its global fraud command centre. The numbers that the payments giant grapples with are enormous. Every year, £15 trillion flows through Visa's networks, representing roughly 15 per cent of the world's economy. And bad actors constantly try to siphon off some of that money. Modern fraudsters vary dramatically in sophistication. To stay ahead, Visa has invested £12 billion over the past five years building AI-powered cyber fraud detection capabilities, knowing that criminals are also spending big. "You have everybody from a single individual threat actor looking to make a quick buck all the way to really corporatised criminal organisations that generate tens or hundreds of millions of pounds annually from fraud and scam activities," Michael Jabbara, Visa's global head of fraud solutions, told AFP during a tour of the company's security campus. "These organisations are very structured in how they operate." The best-resourced criminal syndicates now focus on scams that directly target consumers, enticing them into purchases or transactions by manipulating their emotions. "Consumers are continuously vulnerable. They can be exploited, and that's where we've seen a much higher incidence of attacks recently," Jabbara said. Scam centres The warning signs are clear: anything that seems too good to be true online is suspicious and romance opportunities with strangers from distant countries are especially dangerous. "What you don't realise is that the person you're chatting with is more likely than not in a place like Myanmar," Jabbara warned. He said human-trafficking victims are forced to work in multi-billion-pound cyber scam centres built by Asian crime networks in Myanmar's lawless border regions. The most up-to-date fraud techniques are systematic and quietly devastating. Once criminals obtain your card information, they automatically distribute it across numerous merchant websites that generate small recurring charges — amounts low enough that victims may not notice for months. Some of these operations increasingly resemble legitimate tech companies, offering services and digital products to fraudsters much like Google or Microsoft cater to businesses. On the dark web, criminals can purchase comprehensive fraud toolkits. "You can buy the software. You can buy a tutorial on how to use the software. You can get access to a mule network on the ground or you can get access to a bot network" to carry out denial-of-service attacks that overwhelm servers with traffic, effectively shutting them down. Just as cloud computing lowered barriers for start-ups by eliminating the need to build servers, "the same type of trend has happened in the cyber crime and fraud space," Jabbara explained. These off-the-shelf services can also enable bad actors to launch brute force attacks on an industrial scale — using repeated payment attempts to crack a card's number, expiry date, and security code. The sophistication extends to corporate-style management, Jabbara said. Some criminal organisations now employ chief risk officers who determine operational risk appetite. They might decide that targeting government infrastructure and hospitals generates an excessive amount of attention from law enforcement and is too risky to pursue. 'Millions of attacks' To combat these unprecedented threats, Jabbara leads a payment scam disruption team focused on understanding criminal methodologies. From a small room called the Risk Operations Centre in Virginia, employees analyse data streams on multiple screens, searching for patterns that distinguish fraudulent activity from legitimate credit card use. In the larger Cyber Fusion Centre, staff monitor potential cyberattacks targeting Visa's own infrastructure around the clock. "We deal with millions of attacks across different parts of our network," Jabbara noted, stating that most are handled automatically without human intervention. Visa maintains identical facilities in London and Singapore, ensuring 24-hour global vigilance.
New Straits Times
11 hours ago
- New Straits Times
SpaceX to invest US$2bil in Musk's xAI startup, WSJ reports
NEW YORK: SpaceX has committed US$2 billion to xAI as part of a US$5 billion equity round, deepening the ties between tech billionaire Elon Musk's ventures as his artificial intelligence startup races to compete with rival OpenAI, the Wall Street Journal reported on Saturday. The investment follows xAI's merger with X and values the combined company at US$113 billion, with the Grok chatbot now powering Starlink support and eyed for future integration into Tesla's Optimus robots, the report added. In response to a post on X about whether Tesla, could also invest in xAI, Elon Musk said on Sunday, "It would be great, but subject to board and shareholder approval," without confirming or denying the Journal report on SpaceX's investment plans in xAI. SpaceX, xAI and Tesla did not immediately respond to requests for comment. Reuters could not immediately confirm the WSJ report. Despite recent controversies involving Grok's responses, Musk has called it "the smartest AI in the world," and xAI continues to spend heavily on model training and infrastructure.
Malay Mail
13 hours ago
- Malay Mail
SpaceX pumps US$2b into Elon Musk's xAI, valuing merged firm at US$113b as Grok powers Starlink and eyes Tesla integration
BENGALURU, July 13 — SpaceX has committed US$2 billion to xAI as part of a US$5 billion equity round, deepening the ties between tech billionaire Elon Musk's ventures as his artificial intelligence startup races to compete with rival OpenAI, the Wall Street Journal reported yesterday. The investment follows xAI's merger with X and values the combined company at US$113 billion, with the Grok chatbot now powering Starlink support and eyed for future integration into Tesla's Optimus robots, the report added. SpaceX and xAI did not immediately respond to Reuters' requests for comment. Despite recent controversies involving Grok's responses, Musk has called it 'the smartest AI in the world', and xAI continues to spend heavily on model training and infrastructure.
