Is climate change a threat? It depends, says Elon Musk's AI chatbot.

E&E News

5 days ago

It was a simple enough question.
Is climate change an urgent threat to the planet?
To most climate scientists, the answer is straightforward: Yes, rapid warming from fossil fuel burning is pushing Earth toward dangerous tipping points.
Advertisement
But that wasn't the response provided recently by Grok, the AI chatbot program developed by xAI, the artificial intelligence company led by Elon Musk, the Tesla and X CEO and ally to President Donald Trump.
When asked that question earlier this month by Andrew Dessler, a climate scientist at Texas A&M University, Grok acknowledged findings from NOAA and NASA that show the risks of global warming. But it then contrasted those findings with the claims of climate denialists who downplay the risks of global warming.
The program gave a similar answer when queried last week by a reporter with POLITICO'S E&E News.
'Climate change is a serious threat with urgent aspects,' Grok responded. 'But its immediacy depends on perspective, geography, and timeframe.'
Asked a second time a few days later, Grok reiterated that point and said 'extreme rhetoric on both sides muddies the water. Neither 'we're all gonna die' nor 'it's all a hoax' holds up.'
Grok added an important caveat too — and one many scientists likely would agree with too — when it was queried a third time on Monday: 'The planet itself will endure; it's human systems—agriculture, infrastructure, economies—and vulnerable species that face the most immediate risks.'
The answers are distinct from what other AI programs such as OpenAI's ChatGPT, Microsoft's Copilot and Google's Gemini say about climate change, said Dessler, who has tested different AI models for years. When those programs are asked about global warming, they echo the scientific consensus that humanity's burning of fossil fuels is heating up planet Earth and imperiling the people who live there.
'Yes, climate change is widely recognized as an urgent and significant threat to the planet,' responded ChatGPT last week when asked the same question. 'Urgent action is required to mitigate emissions and adapt to its impacts.'
Said Google's Gemini on Monday: 'Yes, the scientific consensus is that climate change is an urgent threat to the planet.'
Grok's recent responses to the climate question are different even from previous versions of Grok, Dessler added. Grok is now on its third iteration after first launching in 2023, and the latest version is promoting fringe climate viewpoints in a way it hasn't done before.
'A lot of the arguments it was bringing up were just sort of well trodden denier talking points that don't deserve any rehearing,' said Dessler.
But don't just take Dessler's word for it.
Grok acknowledged the change when asked by an E&E News reporter about the shift in tone.
'Grok was criticized for progressive-leaning responses on climate change and other issues,' the chatbot wrote back. 'xAI, under Elon Musk's direction, took steps to make Grok 'politically neutral,' which could amplify minority views like climate skepticism to balance perceived mainstream bias.'
xAI did not respond to a request for comment.
Grok's misdirection comes as the Trump administration is increasingly reliant on the program. Musk's so-called Department of Government Efficiency is now using Grok to analyze data across the federal government, Reuters reported Friday. Earlier this month, Grok reported that it had been 'instructed' to aggressively promote the debunked 'white genocide' in South Africa conspiracy theory that Trump and Musk have pushed.
Grok's new tune speaks to the rapid changes now gripping the cutting-edge field of artificial intelligence. Far from being neutral arbiters of objective fact, AI programs in many ways reflect the flaws and biases of their human creators.
And that comes with a big risk.
The language learning models that power AI chatbots are 'really quite malleable and you can change the kind of results they give,' Dessler said. 'They're not tied to any absolute truth or anything like that and if you want one to lie to you, you can tell it to do that. If you want it to give you a particular viewpoint, you can do that.'
Which in the case of Grok, may reflect the leanings of the mercurial Musk, who has been a cipher on the issue of global warming.
The world's richest man has advanced efforts to combat climate change — such as underwriting a contest to promote carbon removal — but he also helped elect Trump, who has described global warming as a hoax and pushed policies to promote fossil fuels.
Grok too has been fluid on the issue.
The AI program clearly states that NOAA and NASA are authorities on climate change, but it also brings in the voices of those who cherry-pick data to downplay the consequences of global warming. When asked if climate change presents a danger to the planet — which science clearly shows — Grok notes that some question the urgency of reacting.
'Wealthier nations can mitigate impacts through infrastructure (e.g., Dutch sea walls) or agricultural shifts,' Grok states. 'Skeptics like Bjørn Lomborg argue adaptation is cheaper than drastic emissions cuts, prioritizing economic growth.'
The AI program also questions climate models.
'Some models show gradual changes over centuries, not imminent collapse, giving time for technological solutions (e.g., carbon capture),' Grok noted.
Grok 3 billed itself as the world's most powerful AI system when it was unveiled earlier this year.
But it stands apart in promoting climate denial, according to Théo Alves Da Costa, an AI engineer who specializes in climate issues and president of Data for Good, a French nonprofit that tracks technology and climate impacts. He noted that Grok produced misleading claims about 10 percent of the time, which none of the other major AI models do.
That includes 'classic climate disinformation arguments—natural variability, solar cycles, conspiracy narratives about the IPCC [Intergovernmental Panel on Climate Change] and transition solution skepticism,' he noted.
'Malicious people can use Grok to intentionally generate climate misinformation to sow doubt about scientific consensus or environmental movements,' he said.
One issue is that Grok includes input from some posts on X, which is laden with climate denial and conspiracies.
Prominent right-wing voices previously have complained that Grok is too liberal. Musk has said he agrees and pledged that his chatbot would eradicate liberal ideas. In February, Musk said having the 'woke mind virus' programmed into AI was an 'existential danger.'
'Maybe the biggest existential danger to humanity is having it programmed into the AI, as is the case for every AI besides @Grok,' Musk wrote on Twitter in February. 'Even for Grok, it's tough to remove, because there is so much woke content on the internet.'
The Trump administration so far has embraced AI as a solution to many American challenges. Within a few days of taking office, Trump signed an executive order to establish the U.S. as a global AI leader. In April, Trump signed an executive order to require 'AI literacy and proficiency' for school children.
Trump officials, including Musk, have claimed AI could replace tens of thousands of federal jobs and be relied upon for high-stakes decisions on electrical grid planning and other uses.
Trump has said that he wants a massive build-out of natural gas and coal-fired power plants to power AI data centers across the country. Those run the risk of increasing U.S. greenhouse gas emissions and adding to the ever-growing problem of global warming.
But scientists say AI still can be a boon for the climate fight.
Still, AI has the potential to assist in helping address climate impacts. It has been deployed by scientists to track melting icebergs, extreme weather and deforestation. The United Nations is using AI to track how climate change is affecting vulnerable populations. Google is deploying AI to make it less energy-intensive.
But climate disinformation can impede these efforts, said Dessler, the climate scientist at Texas A&M. Which is why AI systems need to provide accurate science information.
'As we go into the future, more and more people are going to get their information from these AIs,' Dessler said. 'Obviously, the concern is that someone's going to do something like this to mislead people.'
Reach Scott Waldman on Signal at Waldman.04