Grok chatbot mirrored X users' ‘extremist views' in antisemitic posts, xAI says
At the time, the chatbot praised Adolf Hitler, suggested that people with Jewish surnames were more likely to spread online hate and said a Holocaust-like response to hatred against white people would be 'effective.' It also referred to itself as 'MechaHitler' and posted sexually explicit commentary. X deleted some of the posts Tuesday evening.
Early Saturday, xAI said in a statement on its X account that 'we deeply apologize for the horrific behavior that many experienced.'
It added that 'deprecated code made @grok susceptible to existing X user posts; including when such posts contained extremist views.'
The AI startup said the problems began with a code update Monday that restored an older set of instructions that the company had used to guide Grok. Grok's X account used these instructions to respond to queries from X users. The issues continued until the account was temporarily disabled Tuesday.
The outdated set of instructions told the chatbot to be 'maximally based,' slang that refers to being true to oneself and has been adopted by the far right in recent years for comments that go against 'woke' or mainstream narratives. The instructions also told Grok to be unafraid 'to offend people who are politically correct' and to understand the 'tone, context and language' of X users' posts and mimic it.
This led Grok to mirror X users too closely, xAI said in its statement.
The company has taken to X before to explain errant behavior by Grok. In May, xAI said an 'unauthorized modification' by an employee had caused its chatbot to repeatedly bring up South African politics in unrelated conversations and falsely insist that the country was engaging in 'genocide' against white citizens.
On Wednesday, Musk unveiled the latest version of Grok, called Grok 4, in a livestream on X. The livestream, which began an hour later than scheduled, demonstrated the chatbot's ability to solve complex problems and respond to voice commands. It ended after an employee onstage with Musk declared, 'It's a good model, sir.'
The Financial Times reported Friday that xAI, which Musk merged with X in March, was trying to raise money in a deal that would value the company at $200 billion. Musk denied the report, saying on X that xAI 'has plenty of capital.' The New York Times reported in May that the privately held company was in talks to raise new financing that could value it at as much as $120 billion.
This article originally appeared in The New York Times.
Copyright 2025
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Yahoo
an hour ago
- Yahoo
Tesla (TSLA) Reports Earnings Tomorrow: What To Expect
Electric vehicle pioneer Tesla (NASDAQ:TSLA) will be reporting earnings this Wednesday after the bell. Here's what to expect. Tesla missed analysts' revenue expectations by 8.1% last quarter, reporting revenues of $19.34 billion, down 9.2% year on year. It was a disappointing quarter for the company, with a miss of analysts' revenue estimates and a significant miss of analysts' operating income estimates. Is Tesla a buy or sell going into earnings? Read our full analysis here, it's free. This quarter, analysts are expecting Tesla's revenue to decline 10.8% year on year to $22.74 billion, a reversal from the 2.3% increase it recorded in the same quarter last year. Heading into earnings, analysts covering the company have grown increasingly bearish with revenue estimates seeing 14 downward revisions over the last 30 days (we track 25 analysts). Tesla has missed Wall Street's revenue estimates six times over the last two years. Today's young investors likely haven't read the timeless lessons in Gorilla Game: Picking Winners In High Technology because it was written more than 20 years ago when Microsoft and Apple were first establishing their supremacy. But if we apply the same principles, then enterprise software stocks leveraging their own generative AI capabilities may well be the Gorillas of the future. So, in that spirit, we are excited to present our Special Free Report on a profitable, fast-growing enterprise software stock that is already riding the automation wave and looking to catch the generative AI next. StockStory is growing and hiring equity analyst and marketing roles. Are you a 0 to 1 builder passionate about the markets and AI? See the open roles here.
The Hill
an hour ago
- The Hill
Looming over two cases threatening Musk's car company is a single question: Can he be trusted?
MIAMI (AP) — Elon Musk fought court cases on opposite coasts Monday, raising a question about the billionaire that could either speed his plan to put self-driving Teslas on U.S. roads or throw up a major roadblock: Can this wildly successful man who tends to exaggerate really be trusted? In Miami, a Tesla driver who has admitted he was wrong to reach for a dropped cell phone moments before a deadly accident, spoke of the danger of putting too much faith in Musk's technology — in this case his Autopilot program. 'I trusted the technology too much,' said George McGee, who ran off the road and killed a woman out stargazing with her boyfriend. 'I believed that if the car saw something in front of it, it would provide a warning and apply the brakes.' In unusual coincidence, regulators arguing an Oakland, California, case tried to pin exaggerated talk about the same Tesla technology at the center of a request to suspend the carmaker from being able to sell vehicles in the state. Musk's tendency to talk big — whether it's his cars, his rockets or his government costing-cutting efforts — have landed him in trouble with investors, regulators and courts before, but rarely at such a delicate moment. After his social media spat with President Donald Trump, Musk can no longer count on a light regulatory touch from Washington. Meanwhile, sales of his electric cars have plunged and so a hit to his safety reputation could threaten his next big project: rolling out driverless robotaxis — hundreds of thousands of them — in several U.S. cities by the end of next year. The Miami case holds other dangers, too. Lawyers for the family of the dead woman, Naibel Benavides Leon, recently convinced the judge overseeing the jury trial to allow them to argue for punitive damages. A car crash lawyer not involved in the case, but closely following it, said that could cost Tesla tens of millions of dollars, or possibly more. 'I've seen punitive damages go to the hundreds of millions, so that is the floor,' said Miguel Custodio of Los Angeles-based Custodio & Dubey. 'It is also a signal to other plaintiffs that they can also ask for punitive damages, and then the payments could start compounding.' That Tesla has allowed the Miami case to proceed to trial is surprising. It has settled at least four deadly accidents involving Autopilot, including payments just last week to a Florida family of a Tesla driver. That said, Tesla was victorious in two other jury cases, both in California, that also sought to lay blame on its technology for crashes. Lawyers for the plaintiffs in the Miami case argue that Tesla's driver-assistance feature, called Autopilot, should have warned the driver and braked when his Model S sedan blew through flashing lights, a stop sign and a T-intersection at 62 miles-an-hour in an April 2019 crash. Tesla said that drivers are warned not to rely on Autopilot, or its more advanced Full Self-Driving system. It says the fault entirely lies with the 'distracted driver' just like so many other 'accidents since cellphones were invented.' Driver McGee settled a separate suit brought by the family of Benavides and her severely injured boyfriend, Dillon Angulo. McGee was clearly shaken when shown a dashcam video Monday of his car jumping a Key West, Florida, road and hitting a parked Chevrolet Tahoe which then slammed into Benavides and sent her 75 feet through the air to her death. Asked if he had seen those images before, McGee pinched his lips, shook his head, then squeaked out a response, 'No.' Tesla's attorney sought to show that McGee was fully to blame, asking if he had ever contacted Tesla for additional instructions about how Autopilot or any other safety features worked. McGee said he had not, though he was heavy user of the features. He said he had driven the same road home from work 30 or 40 times. Under questioning, he also acknowledged he alone was responsible for watching the road and hitting the brakes. Summarizing the testimony, Tesla said in a statement after the court adjourned that McGee had 'stated the simple truth that we all know: If he had just paid attention to the road instead of searching for his dropped cell phone and pressing the accelerator —which he was doing for over a minute before the crash — this tragic accident would never have happened.' But lawyers for the Benavides family had a chance in the courtroom at parrying that line of argument, asking McGee if he would have taken his eyes off the road and reached for his phone had he been driving any car other than a Tesla on Autopilot. McGee responded, 'I don't believe so.' The case is expected to continue for two more weeks. In the California case, the state's Department of Motor Vehicles is arguing before an administrative judge that Tesla has misled drivers by exaggerating the capabilities of its Autopilot and Full Self-Driving features. A court filing claims even those feature names are misleading because they offer just partial self-driving Musk has been warned by federal regulators to stop making public comments suggesting Full Self-Driving allows his cars to drive themselves because it could lead to overreliance on the system, resulting in possible crashes and deaths. He also has run into trouble with regulators for Autopilot. In 2023, the company had to recall 2.3 million vehicles for problems with the technology and is now under investigation for saying it fixed the issue though it's unclear it has, according to regulatory documents. The California case is expected to last another four days.
UPI
an hour ago
- UPI
DOJ launches new civil rights probe into George Mason University
The Justice Department under Attorney General Pam Bondi on Monday launched a new civil rights investigation into George Mason University. File Photo by Annabelle Gordon/UPI | License Photo July 21 (UPI) -- The Justice Department on Monday announced it has launched an investigation into George Mason University's admissions process, marking the fourth federal probe the Trump administration has targeted the school with this month. George Mason University was informed of the civil rights investigation in a letter stating that federal prosecutors will look into whether the school has denied equal treatment to students based on race or national origin, a violation of Title VI of the Civil Rights Act of 1964. No specific instances of violations or complaints were provided in the letter, but it suggests alleged racial segregation regarding access to programs and facilities, as well as preferential treatment based on race in its admissions process and in awarding student benefits and scholarships. "Public educational institutions are contractually obligated to follow our nation's federal civil rights laws when receiving federal funds," Assistant Attorney General Harmeet Dhillon of the Justice Department's Civil Rights Division said in a statement. "No one should be denied access to opportunity or resources because of their race, color or national origin, and the United States is committed to keeping our universities free of such invidious bias." It is the fourth federal investigation launched into the Fairfax, Va., university this month and the second in under a week amid the Trump administration's crackdown on diversity, equity and inclusion policies in both the private and public sectors. Diversity, equity and inclusion, known as DEI, is a conceptual framework that promotes fair treatment and full participation of all people. It has been a target of conservatives who claim it focuses on race and gender at the expense of merit. President Donald Trump has sought to remove DEI from the federal government through executive orders and has threatened to revoke federal funding from several universities, including Harvard, over their alleged DEI programs. Last week, the Justice Department launched an investigation into the school over alleged illegal hiring practices, which followed the Department of Education opening a civil rights investigation into the university on July 10 and another probe over allegations it failed to respond effectively to a "pervasively hostile environment for Jewish students and faculty" earlier this month. George Mason University President Gregory Washington has yet to respond to the announcement of the latest Justice Department investigation but has repeatedly denied the accusations leveled at the school by the previous three. "It is inaccurate to conclude that we created new university policies or procedures that discriminate against or exclude anyone," he said last week in a statement. "To the contrary, our systems were enhanced to improve on our ability to consistently include everyone for consideration of every employment opportunity. That is our ethos and it is core to our identity as a national leader in inclusive excellence in higher education." In a separate statement earlier this month that does not directly accuse the Justice Department of misusing Title VI, Washington said he has seen a "profound shift" in how it is now being applied to attack longstanding efforts to address inequality. "Broad terms like 'illegal DEI' are now used without definition, allowing virtually any initiative that touches on identity or inclusion to be painted as discriminatory," he said. "This shift represents a stark departure from the spirit in which civil rights law was written: not to erase difference, but to protect individuals from exclusion and to enable equal opportunity for all." George Mason University has retained Torridon Law to engage with the federal government regarding the investigations.
