Grok shows ‘flaws' in fact-checking Israel-Iran war: study
WASHINGTON: Elon Musk's AI chatbot Grok produced inaccurate and contradictory responses when users sought to fact-check the Israel-Iran conflict, a study said Tuesday, raising fresh doubts about its reliability as a debunking tool.
With tech platforms reducing their reliance on human fact-checkers, users are increasingly utilizing AI-powered chatbots — including xAI's Grok — in search of reliable information, but their responses are often themselves prone to misinformation.
'The investigation into Grok's performance during the first days of the Israel-Iran conflict exposes significant flaws and limitations in the AI chatbot's ability to provide accurate, reliable, and consistent information during times of crisis,' said the study from the Digital Forensic Research Lab (DFRLab) of the Atlantic Council, an American think tank.
'Grok demonstrated that it struggles with verifying already-confirmed facts, analyzing fake visuals, and avoiding unsubstantiated claims.'
The DFRLab analyzed around 130,000 posts in various languages on the platform X, where the AI assistant is built in, to find that Grok was 'struggling to authenticate AI-generated media.'
Following Iran's retaliatory strikes on Israel, Grok offered vastly different responses to similar prompts about an AI-generated video of a destroyed airport that amassed millions of views on X, the study found.
It oscillated — sometimes within the same minute — between denying the airport's destruction and confirming it had been damaged by strikes, the study said.
In some responses, Grok cited the a missile launched by Yemeni rebels as the source of the damage. In others, it wrongly identified the AI-generated airport as one in Beirut, Gaza, or Tehran.
When users shared another AI-generated video depicting buildings collapsing after an alleged Iranian strike on Tel Aviv, Grok responded that it appeared to be real, the study said.
The Israel-Iran conflict, which led to US air strikes against Tehran's nuclear program over the weekend, has churned out an avalanche of online misinformation including AI-generated videos and war visuals recycled from other conflicts.
AI chatbots also amplified falsehoods.
As the Israel-Iran war intensified, false claims spread across social media that China had dispatched military cargo planes to Tehran to offer its support.
When users asked the AI-operated X accounts of AI companies Perplexity and Grok about its validity, both wrongly responded that the claims were true, according to disinformation watchdog NewsGuard.
Researchers say Grok has previously made errors verifying information related to crises such as the recent India-Pakistan conflict and anti-immigration protests in Los Angeles.
Last month, Grok was under renewed scrutiny for inserting 'white genocide' in South Africa, a far-right conspiracy theory, into unrelated queries.
Musk's startup xAI blamed an 'unauthorized modification' for the unsolicited response.
Musk, a South African-born billionaire, has previously peddled the unfounded claim that South Africa's leaders were 'openly pushing for genocide' of white people.
Musk himself blasted Grok after it cited Media Matters — a liberal media watchdog he has targeted in multiple lawsuits — as a source in some of its responses about misinformation.
'Shame on you, Grok,' Musk wrote on X. 'Your sourcing is terrible.'
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Arab News
2 hours ago
- Arab News
Comfort or isolation: Pakistanis weigh pros and cons of ChatGPT as confidant
LAHORE: When Mehak Rashid looks back on a restless, emotionally fragile phase of her life earlier this year, an unlikely confidant comes to mind. 'When nobody else was listening to you and everybody else thought you were crazy, ChatGPT was there,' Rashid, a metallurgy and materials engineer from Lahore, told Arab News. 'I just wanted to be heard… It will not give you a judgment and that's so beautiful.' Rashid began using the chatbot after noticing her children experimenting with it for schoolwork. Now, she often turns to it for 'answers' and 'different perspectives.' 'It helps me in every way,' she said. Since its launch in November 2022, ChatGPT has attracted hundreds of millions of users and, by mid-2025, logged nearly 800 million weekly active users. Many in Pakistan, among the top 20 countries for ChatGPT traffic, use it daily for emotional support, venting feelings, or late-night reassurance when friends aren't available. Globally, an estimated 40 percent of ChatGPT conversations relate to mental well-being, and a Sentio University survey found nearly half of users with ongoing mental health issues rely on it for support: 73 percent for anxiety, 63 percent for advice, and 60 percent for help with depression. While this instant comfort helps some cope, psychologists warn that heavy reliance on AI can weaken real human connections and deepen social isolation in a country already short on mental health resources. A March 2025 study by OpenAI and MIT found frequent users reported increased dependence and loneliness, suggesting that AI companionship can erode human bonds and intensify feelings of isolation rather than resolve them. For Lahore-based designer Khizer Iftikhar, ChatGPT began as a professional aid but gradually crept into his personal life and started affecting his relationships, especially with his wife. 'I have a very avoidant attachment style,' he said. 'Instead of confronting someone, I can just talk about the good part with people and let the chatbots handle the negative part.' Iftikhar described ChatGPT as 'a multiple personality tool' that lacked the balance of real human interaction. Many experts say using AI models can weaken bonds overtime, reduce empathy, and make people more emotionally self-contained, preferring the predictable reassurance of a machine over the give-and-take of genuine human connection. 'With humans, relationships are about give and take. With chatbots, it's not like that,' Iftikhar said. Despite once trying therapy, he now uses ChatGPT to process emotions and trusts people only for practical advice. 'I would trust a chatbot more when it comes to the feelings part,' Iftikhar said. 'But when it comes to the work part, I will trust humans more.' In Islamabad, 26-year-old Tehreem Ahmed initially used ChatGPT for office transcriptions and calorie tracking but it eventually became an emotional lifeline. One night, overwhelmed by troubling news and unable to reach friends, she turned to the chatbot. 'It was around 3am and none of my friends were awake,' she said. 'So, I went on ChatGPT and I typed in all that I got.' The chatbot encouraged her to pause and reflect before reacting. 'I feel like it responded well because I gave it a smarter prompt… Had I just said, 'Hey, this has happened. What should I do?' I guess it would have just given me all the options… I could have self-sabotaged.' While Ahmed doesn't fully trust the bot, she said she preferred it to people who might dismiss her feelings. 'If I know my friend is not going to validate me, I'd rather go to the bot first.' 'DETERIORATING HUMAN CONNECTIONS' For one anonymous Lahore-based tech professional, ChatGPT quickly shifted from a practical helper to an emotional crutch during a difficult relationship and the ongoing war in Gaza. She first used it in late 2023 to navigate a job change, edit CVs, and prepare for assessments. But emotional upheaval deepened her reliance on the bot. 'That [romantic] relationship didn't progress,' she said. 'And the platform helped me a lot emotionally in navigating it.' Her sessions became so layered and spiritual that some ended in 'prostration from spiritual overwhelm.' Still, she was careful not to project too much onto the tool: 'It's a mirror of my flawed self… I try not to let the tool simply reflect my ego.' Psychologists caution that without the challenges and messiness of real interactions, people using chatbots may lose vital social skills and drift further into isolation. Mahnoor Khan, who runs MSK Clinics in Islamabad, agreed, saying the search for emotional safety in AI was becoming increasingly common as people feared judgment from others. 'Over a period of time, human connections have deteriorated,' the psychologist said. 'When people share something vulnerable with a friend, they often feel judged or lectured.' To avoid that, many turn to chatbots. But Khan warned that AI's constant affirmation could have unintended consequences. 'It will tell you what you want to listen to… If you're happy, it's your companion; if you're sad, it instantly talks to you. The downside is that you are getting away from socialization.' The trend is especially troubling in a country where mental health care remains deeply under-resourced: Pakistan has fewer than 500 psychiatrists for a population of over 240 million, according to WHO estimates. No wonder then that even people with clinical mental health issues were turning to AI. Khan recalled the case of a young woman who used ChatGPT so often that it replaced nearly all her social interaction. 'She had a lot of suicidal ideations,' Khan said. 'She kept feeding ChatGPT: 'I feel very depressed today… you tell me what I should do?' ChatGPT kept telling her to avoid friends like that.' Eventually, she cut everyone off. One day, she asked the chatbot what would happen if she overdosed on phenyl. 'ChatGPT said, 'There are no consequences. In case you overdose yourself, you might get paralyzed,'' Khan recalled. The girl only read the first half and attempted suicide. She survived.
Arab News
9 hours ago
- Arab News
Grok shows ‘flaws' in fact-checking Israel-Iran war: study
WASHINGTON: Elon Musk's AI chatbot Grok produced inaccurate and contradictory responses when users sought to fact-check the Israel-Iran conflict, a study said Tuesday, raising fresh doubts about its reliability as a debunking tool. With tech platforms reducing their reliance on human fact-checkers, users are increasingly utilizing AI-powered chatbots — including xAI's Grok — in search of reliable information, but their responses are often themselves prone to misinformation. 'The investigation into Grok's performance during the first days of the Israel-Iran conflict exposes significant flaws and limitations in the AI chatbot's ability to provide accurate, reliable, and consistent information during times of crisis,' said the study from the Digital Forensic Research Lab (DFRLab) of the Atlantic Council, an American think tank. 'Grok demonstrated that it struggles with verifying already-confirmed facts, analyzing fake visuals, and avoiding unsubstantiated claims.' The DFRLab analyzed around 130,000 posts in various languages on the platform X, where the AI assistant is built in, to find that Grok was 'struggling to authenticate AI-generated media.' Following Iran's retaliatory strikes on Israel, Grok offered vastly different responses to similar prompts about an AI-generated video of a destroyed airport that amassed millions of views on X, the study found. It oscillated — sometimes within the same minute — between denying the airport's destruction and confirming it had been damaged by strikes, the study said. In some responses, Grok cited the a missile launched by Yemeni rebels as the source of the damage. In others, it wrongly identified the AI-generated airport as one in Beirut, Gaza, or Tehran. When users shared another AI-generated video depicting buildings collapsing after an alleged Iranian strike on Tel Aviv, Grok responded that it appeared to be real, the study said. The Israel-Iran conflict, which led to US air strikes against Tehran's nuclear program over the weekend, has churned out an avalanche of online misinformation including AI-generated videos and war visuals recycled from other conflicts. AI chatbots also amplified falsehoods. As the Israel-Iran war intensified, false claims spread across social media that China had dispatched military cargo planes to Tehran to offer its support. When users asked the AI-operated X accounts of AI companies Perplexity and Grok about its validity, both wrongly responded that the claims were true, according to disinformation watchdog NewsGuard. Researchers say Grok has previously made errors verifying information related to crises such as the recent India-Pakistan conflict and anti-immigration protests in Los Angeles. Last month, Grok was under renewed scrutiny for inserting 'white genocide' in South Africa, a far-right conspiracy theory, into unrelated queries. Musk's startup xAI blamed an 'unauthorized modification' for the unsolicited response. Musk, a South African-born billionaire, has previously peddled the unfounded claim that South Africa's leaders were 'openly pushing for genocide' of white people. Musk himself blasted Grok after it cited Media Matters — a liberal media watchdog he has targeted in multiple lawsuits — as a source in some of its responses about misinformation. 'Shame on you, Grok,' Musk wrote on X. 'Your sourcing is terrible.'
Al Arabiya
10 hours ago
- Al Arabiya
Musk's 'Robotaxis' Draw Regulatory Scrutiny After Video Shows One Driving in an Opposing Lane
Federal traffic safety regulators are looking into suspected problems with Elon Musk's test run of self-driving robotaxis in Texas after videos surfaced showing them braking suddenly or going straight through an intersection from a turning lane and driving down the wrong side of the road. The National Highway Traffic Safety Administration said Tuesday that it has asked Tesla for information about the apparent errors. Though many other videos show robotaxis driving perfectly, if regulators find any major issues, that would likely raise questions about Musk's repeated statements that the robotaxis are safe and his claim that Tesla will dominate a future in which nearly all cars on the road will have no one behind the wheel–or even need a steering wheel at all. 'NHTSA is aware of the referenced incidents and is in contact with the manufacturer to gather additional information,' the agency said in a statement. Passengers in Tesla robotaxis on the road in Austin, Texas, have generally been impressed, and the stock rose 8 percent Monday. Investors grew more cautious Tuesday after news of NHTSA's inquiry, and the stock fell more than 2 percent. Tesla did not immediately respond to a request for comment. A bullish Tesla financial analyst who was driven around in a robotaxi on Sunday when the test runs began said his ride was 'perfect' and suggested the videos on X and YouTube showing errors were no big deal. 'Any issues they encounter will be fixed,' said Wedbush Securities' Dan Ives, calling the test 'a huge success in the past three days despite the skeptics.' One of those skeptics, a Telemetry Insight expert in car technology, said the videos were alarming enough that the tests, as currently run, should be halted. 'The system has always had highly erratic performance, working really well a lot of the time, but frequently making random and inconsistent, but dangerous errors,' said Sam Abuelsamid in a text referring to Tesla's self-driving software. 'This is not a system that should be carrying members of the public or being tested on public roads without trained test drivers behind the wheel.' In one video, a Tesla moves into a lane with a big yellow arrow indicating it is for left turns only, but then goes straight through the intersection instead, entering an opposing lane on the other side. The car seems to realize it made some mistake and begins to swerve several times, with the steering wheel jerking back and forth, before eventually settling down. But the Tesla proceeds in the opposing lane for 10 seconds. At the time, there was no oncoming traffic. The passenger in the car who posted the video, money manager Rob Maurer, shrugged off the incident. 'There are no vehicles anywhere in sight, so this wasn't a safety issue,' Maurer said in commentary accompanying his video. 'I didn't feel uncomfortable in the situation.' Another video shows a Tesla stopping twice suddenly in the middle of the road, possibly responding to the flashing lights of police cars. But the police are obviously not interested in the Tesla or traffic in front or behind it because they have parked on side roads, not near it, apparently responding to an unrelated event. Federal regulators opened an investigation last year into how Teslas with what Musk calls 'Full Self-Driving' have responded in low-visibility conditions after several accidents, including one that was fatal. Tesla was forced to recall 2.4 million of its vehicles at the time. Musk has said his Teslas using 'Full Self-Driving' are safer than human drivers, and his robotaxis, using a newer, improved version of the system, will be so successful so quickly that he will be able to deploy hundreds of thousands of them on the road by the end of next year. Even if the Austin test goes well, though, the billionaire faces big challenges. Other self-driving companies have launched taxis, including Amazon's Zoox and current market leader Waymo, which is not only picking up passengers in Austin, but several other cities. The company recently announced it had clocked its 10 millionth paid ride. Musk needs a win in robotaxis. His work in the Trump administration as cost-cutting czar has alienated many buyers among Tesla's traditional environmentally conscious and liberal base in the US, tanking sales. Buyers in Europe have balked, too, after Musk embraced some extreme right-wing politicians earlier this year in both Britain and Germany.
