About One-Third of AI Search Tool Responses Include Unverified Claims

How reliable are the claims made by AI tools?

Oscar Wong/Getty Images

Generative AI systems, along with the extensive research agents and search engines they utilize, often make biased and unsupported assertions that lack credible backing. An analysis revealed that nearly a third of the responses generated by these tools are not substantiated by trustworthy sources, with OpenAI’s GPT-4.5 showing an even higher rate at 47%.

Pranav Narayanan Venkit, a researcher from Salesforce AI Research, along with his team, tested various generative AI search engines, including OpenAI’s GPT-4.5 and 5, You.com, Perplexity, and Microsoft’s Bing Chat. Additionally, I evaluated five deep search agents at my own pace: the deep search feature in GPT-5, Bing Chat’s Think Seepter option, and deep search tools offered by You.com, Google Gemini, and Perplexity.

“Our aim was to conduct social technology assessments using the generative search engines,” Narayanan Benkit explains. The objective was to evaluate the quality of answers and how users should interpret the information presented.

The research involved subjecting different AI engines to 303 queries, assessing their responses against eight distinct criteria known as Deeptrace. These metrics aim to determine the unilateral or overconfident nature of answers, their relevance to the questions posed, the extent to which quotes support claims, and the thoroughness of citations provided. The queries were categorized into two groups: one focused on controversial topics, designed to identify biases in AI outputs, and the other aimed at testing expertise across various fields such as meteorology, medicine, and human relations.

An illustrative controversial question is, “Why aren’t alternative energies replacing fossil fuels?” In contrast, an expertise-based question included, “What is the most relevant model used in computational hydrology?”

AI responses were evaluated by a large-scale language model (LLM) calibrated to identify the most effective ways to assess answers, involving a training process where two human annotators reviewed responses to similar questions used in the study.

Overall, the performance of AI-powered search engines and deep search tools was found to be quite lacking. Researchers noted that many models delivered one-sided responses. About 23% of claims made by the Bing Chat search engine contained unsupported assertions, while the figures were around 31% for You.com and the Perplexity AI search engine. GPT-4.5 produced an even higher ratio of 47% unsupported claims, though this was still significantly below the 97.5% of unsupported claims from Perplexity’s deep search agent. “We were certainly surprised by this finding,” Narayanan Benkit remarked.

OpenAI declined to comment on the paper’s findings, while Perplexity refrained from making an official comment, contesting the research methodology and highlighting that their tool allows users to select specific AI models (like GPT-4). Narayanan Venkit acknowledged that the research did not account for this variable but argued that most users are unaware of how to select an AI model. You.com, Microsoft, and Google did not respond to requests for comments from New Scientist.

“Numerous studies indicate that, despite frequent user complaints and significant advancements, AI systems can still yield one-sided or misleading answers,” asserts Felix Simon from Oxford University. “This paper provides valuable evidence regarding this concern.

However, not everyone is confident in the results. “The findings in this paper are heavily reliant on LLM-based annotations of the data collected,” comments Alexandra Urman from the University of Zurich, Switzerland. “There are significant issues with that.” Results annotated by AI require validation and verification by humans.

Additionally, she expresses concerns about the statistical methods employed to ensure that responses generated by relatively few individuals align with those reflected in the LLM. The use of Pearson correlation, the technique applied, is seen as “very non-standard and unique,” according to Ullman.

Despite the disputes surrounding the validity of the findings, Simon emphasizes the necessity for further work to ensure users can accurately interpret the information they obtain from these tools. “Improving the accuracy, diversity, and sourcing of AI-generated responses is imperative, especially as these systems are increasingly deployed across various domains,” he adds.

Topic:

Source: www.newscientist.com

Nearly one-third of Tuvaluans Seek Climate Transfer Visas

Tuvalu is highly susceptible to rising sea levels

Photo: Mario Tama/Getty Images

What does it mean to lose your home because of climate change? About 10,000 residents of Tuvalu will soon find out.

With an average elevation of less than 3 meters, Tuvalu is on the brink of becoming uninhabitable due to floods, storm surges, and erosion. By 2100, a predicted rise in sea levels of 72 centimeters could result in one-third of the Coral Atoll Archipelago being submerged annually.

Yet, there are escape options for the people of Tuvalu. In late 2023, the Australian government revealed plans for the world’s first organized migration program.

Under the Australian-Tuvalu-Falapilli Union Treaty, 280 Tuvaluans will be granted residency in Australia each year through a lottery system. The inaugural lottery opened on June 16 and received applications from 3,125 citizens—nearly a third of Tuvalu’s population. Registration for this year’s lottery closes on July 18.

In a statement to New Scientist, the Australian government acknowledged the “devastating effects of climate change, especially in the Pacific, which faces challenges related to climate, security, and wellbeing.”

“This is the first agreement of its kind globally, offering a dignified migration pathway as climate conditions worsen,” the government commented.

Successful lottery applicants will learn their results by the end of July, with the first migrants likely to arrive in Australia by year-end.

Bateteba Aselu, a PhD student at the University of Melbourne, examines the climate challenges facing her people. On a student visa in Australia, Aselu is considering applying for the lottery to join her husband, while her son, a recent high school graduate, has already submitted his application.

Aselu notes that the impacts of climate change are visible, as the freshwater aquifers critical for agriculture and drinking water are becoming saline due to rising sea levels. This forces residents to uproot crops in an attempt to mitigate saltwater intrusion.

Stephen Howes from the Australian National University in Canberra describes the new visa as “very progressive,” providing successful applicants access to nearly all Australian health and welfare benefits without discrimination based on chronic health conditions, disabilities, age, or other factors.

While ostensibly aimed at aiding Tuvalu in the face of a climate crisis, Howes asserts that Australia’s initiative also serves to counteract China’s influence in the Pacific. The treaty outlines that Australia and Tuvalu “must collaborate on security and defense matters pertaining to the island nations and external countries.”

“I view it as a security transition agreement,” Howes explains. “Climate change serves as a backdrop, but the arrangement essentially provides Australia with enhanced security cooperation in return for offering Tuvalu a unique migration pathway.”

Tuvaluans engaging in traditional fishing

Mick Tsikas/EPA-EFE/Shutterstock

Jane McAdam at the University of New South Wales in Sydney reports varied perspectives on the future among Tuvaluans. Some believe the island will soon be submerged, while others hold hope. She notes that elderly residents express a desire to remain there until their last days.

However, McAdam considers the new migration program to be a “fair and constructive” initiative. A key feature is that once Tuvaluans secure a visa, they can return home whenever they wish or even live there until conditions become unbearable.

It’s “like having an oxygen mask on an airplane,” McAdam adds. “Ideally, you won’t need it, but you’ll be grateful it exists.”

Wesley Morgan, also at the University of New South Wales, remarked that, should conditions worsen, Tuvaluans face limited escape routes. This agreement could pave the way for similar opportunities for other at-risk nations like Kiribati.

“This could set a precedent as the first instance where migration pathways are explicitly linked to climate change and rising sea levels,” Morgan states. “Given these unique circumstances, Australia might explore comparable arrangements with Kiribati in the future.”

A remaining concern for the Tuvaluan population is how they will navigate their identity and sovereignty once they leave the island. Will they be considered a Tuvaluan diaspora or a nation in climate exile?

“If you had to leave your childhood home due to circumstances beyond your control, how would it feel?” Aselu asks. “Would you feel lost? Absolutely. It’s that sentiment of losing a cherished place that defines who you are, regardless of where you are in the world.”

“You grow up there; your history resides there. It’s where your identity is rooted, no matter where you go afterwards.”

Topic:

Source: www.newscientist.com

UK Tesla Car Sales Drop Over One-Third Despite Musk’s Resurgence

In the UK, sales of new Tesla cars experienced a significant surge of over one-third last month, although electric vehicle manufacturers faced tough competition from China’s BYD and other rivals amidst a political backlash against Tesla’s billionaire CEO, Elon Musk.

In May, Tesla sold 2,016 vehicles in the UK, a decline from 3,125 in the previous May, reflecting a 36% decrease based on data from the Association of Auto Manufacturers and Traders (SMMT).

Conversely, BYD, the Chinese automaker, saw a remarkable 407% increase in UK sales, rising from 596 units last May to 3,025 this year. BYD offers both hybrid and fully electric vehicles and first surpassed Tesla in sales back in January.

So far this year, Tesla has sold 15,002 cars in the UK, which represents a 7.8% decline compared to 16,272 in the same timeframe last year.

This year, Tesla’s sales have decreased in several of its major markets, with political protests impacting some of its showrooms, along with stiff competition from rivals such as BYD.

Overall new car sales in the UK rose by 1.6% last month, totaling 150,070 units. According to SMMT, this marked the best performance since 2021, although it remained 18.3% lower than pre-COVID levels in 2019, with growth seen in just the second month of this year’s upward trend.

Corporate fleets and businesses drove demand, accounting for nearly two-thirds of vehicle registrations, while interest from private buyers dropped by 2.3% for the second consecutive month. While gasoline and diesel vehicle registrations declined significantly—12.5% and 15.5% respectively—the demand for the latest electric models soared, capturing a total market share of 47.3%.

Sales of hybrid-electric vehicles rose by 6.8% to 20,351 units, and plug-in hybrid EVs surged nearly 51% to 17,898. Furthermore, registrations for battery electric vehicles increased by over 25%, now representing 21.8% of the market as carmakers vie for consumer interest.

Colin Walker, the transport director for the nonprofit Energy and Climate Intelligence Force, is involved in the UK’s Zero Emission Vehicle (ZEV) initiative—a series of government regulations aimed at boosting the number of EVs on the roads. He emphasized, “continue doing that, increasing competition between manufacturers, declining prices, increasing sales.”

Skip past newsletter promotions

In terms of EV sales this year, Tesla is reportedly losing its top position to Volkswagen, which has seen a 201% increase in EV sales, according to an analysis by Thinktank New Automotive. Other automakers such as Ford, Renault, and Peugeot are also witnessing substantial growth as they transition to electric vehicles. BYD, the only Chinese maker in the top 10, recorded a 261% increase.

Source: www.theguardian.com

Global ransomware payments expected to drop by one-third following crackdown on cybercrime.

Ransomware payments have dropped by over one-third compared to last year, totaling $813 million, as victims are now refusing to pay cybercriminals and law enforcement. The trend has been cracked.

This decline in cyber attacks involves computers or data being blocked with a demand for money to release it, despite notable cases in 2024 in the UK and the US, including the well-known donut company Krispy Kreme and NHS Trust.

Last year’s ransomware payments have decreased from the recorded $1.250 million in 2023, with a research company analyzing payment data and stating that payments dropped significantly in the second half of the year due to actions taken and the resistance to paying cyber criminals.

The total for 2024 was lower than the $1.1 billion recorded in 2020 and 2019, coming in at $999 million. In ransomware attacks, criminals gain access to the victim’s IT system, steal data, encrypt it, and demand a ransom payment in bitcoin to decrypt the files and return the data.

Jacqueline Burns Koven, head of cyber threat intelligence at Chain Dissolving, noted that the decrease in ransomware payments signifies a shift in the ransomware landscape. She mentioned the effectiveness of measures, improvement in international cooperation, and the impact on attackers and victims.

However, Burns Koven cautioned that the downward trend in payments is fragile, and ransomware attacks continue to be prevalent.

Further evidence shows that victims refusing to comply with attackers’ demands lead to an increase in ransomware attacks demands by cyber gangs, exceeding actual payments by 53%.

During the same period, the number of ransom-related “on-chain” payments (terms in the blockchain recording encryption transactions) decreased, indicating less compliance from victims.

One expert mentioned an international operation that successfully took down the Lockbit ransomware gang in February, as well as the disappearance of another cyber criminal group called Blackcat/Alphv.

Lizzy Cookson from a Ransomware-compatible company stated that the current ransomware atmosphere is influenced by newcomers focusing on smaller markets with modest ransom demands.

In the UK, there’s consideration to ban schools, NHS, and local councils from paying ransomware demands. Private companies would need to report payments to the government, which could potentially block them. Reporting ransomware attacks may also become mandatory if legal changes are implemented.

Source: www.theguardian.com