AI Skepticism Is on the Rise
https://www.darkreading.com/cybersecurity-operations/contrarians-no-more-ai-skepticism
Publish Date: 2025-12-31 09:00:00
Source Domain: www.darkreading.com
Using an unordered list, summarize the following article with between 4 and 8 key points. As 2025 comes to a close, some of the artificial intelligence industry’s biggest skeptics may be poised for a victory lap.In recent months, AI has taken some hits on several fronts. First and foremost, there are increasing fears about an AI bubble potentially popping as major stocks have dipped. Additionally, several studies have shown many companies have yet to achieve the return on investment they’d hoped for with their generative AI (GenAI) pilots. Meanwhile, threat actors have continued to take advantage of GenAI platforms for malicious activities, and concerns have grown about the technology’s adverse effects on the workforce (not to mention society in general). All of these factors, critics say, have added up to a notable shift in public sentiment — and not in a positive direction.As a result, some of the most well-known AI critics are claiming their predictions have been validated, whether they’re about massive AI spending failing to deliver results or lofty promises about general artificial intelligence falling short. Gary Marcus, emeritus professor of psychology and neural science at NYU and a notable AI skeptic, tells Dark Reading that the shift has underscored one of his long-held beliefs. “Most of the public is NOT enthusiastic about AI; it’s the media and the industry that is enthusiastic,” he says in an email.Related:Dark Reading Confidential: Stop Secrets Creep Across Developer PlatformsSome critics have focused primarily on the economics of the AI boom. Tech entrepreneur and Stanford University lecturer Jerry Kaplan, for example, said in a recent essay that the AI bubble is real, and investments in AI companies are unlikely to keep up with the massive infrastructure spending on data centers and GPUs. “We may be heading for a duet combining the ‘greatest hits’ of the 2000 dot-com bubble and the 2008 housing crisis,” he wrote.Others have focused more on the technical aspects of today’s GenAI platforms. Melanie Mitchell, computer scientist and professor at the Santa Fe Institute, has argued that the industry is overestimating the cognitive and humanlike reasoning capabilities of AI models and that better and more rigorous testing is needed. And whether it’s financial or technical skepticism or both, many enterprises are still searching for the use cases and bottom-line value that the industry has promised. AI Tools: Where’s the Value?A common thread among skeptics and critics is that AI companies have generally overpromised and underdelivered on what their tools can do for businesses. “My own view is that a more sophisticated form of AI, not yet invented, will truly be life-changing, but chatbots aren’t reliable enough to be all that helpful to most people (except, for example, programmers) and that the costs to society are high,” Marcus says.Related:ServiceNow Buys Armis for $7.75B, Boosts ‘AI Control Tower’Rich Mogull, chief analyst at the Cloud Security Alliance, says enterprises have been subjected to pervasive AI marketing since 2022, and some vendors have “rammed down it their throats.” But expected cost savings have yet to materialize.”Enterprises want to get the value out of the technology, but we’re also hearing from them that the providers have to figure out the use cases to get that value — where it’s actually providing value and not just costing more money,” Mogull says.In a recent speech at the University of Washington, Cory Doctorow, author and special adviser to the Electronic Frontier Foundation, argued that AI is stuck in a kind of quagmire. AI coding assistants, for example, can replace entry-level developers (the kind that are paid cheaply on a contract basis), but that won’t generate the kind of value that enterprises truly want.”For AI to be valuable, it has to replace high-wage workers, and those are precisely the experienced workers, with process knowledge, and hard won intuition, who might spot some of those statistically camouflaged AI errors,” Doctorow said in his speech.Related:Afripol Focuses on Regional Cyber Challenges, Deepening CooperationThe senior coders are a key part of the equation because they’re the ones that will spot AI-generated errors, hallucinations, and bugs. But a higher-earning developer, Doctorow argues, is exactly the kind of role that technology executives want to eliminate and replace with AI.Ultimately, Marcus says faith has waned in AI tools largely because industry leaders like Sam Altman and Elon Musk are “constantly overpromising” what AI can do for organizations. “Some people played with the tech and ultimately found they couldn’t count on it,” he says.Cybersecurity Benefits Come at a CostIf there’s been a bright horizon for the AI industry this year, it’s likely been in cybersecurity. Vendors have long embraced machine learning and AI in their products and internal services, but some of the newer applications of the technology have produced compelling results.”If you look at the work done by the AI Cyber Challenge, for example, the auto-identification and patching of vulnerabilities is wild,” Mogull says. “That works, and it’s potentially very impactful.”Still, in its 2025 AI Maturity in Cybersecurity Report, Arkose Labs found that only about half the enterprises it surveyed had realized measurable benefits despite widespread adoption.And many critics have expressed concerns about the high costs of such applications — both to the individual customer and the providers themselves. Additionally, some cybersecurity experts have argued that organizations are overstating the threat of AI-driven cyberattacks to incentivize purchasing AI solutions. For example, noted security researcher Kevin Beaumont recently criticized a now-deleted MIT study that claimed 80% of ransomware attacks were powered by AI.Despite glaring cases of “AI washing” and exaggerated threats, Mogull believes there are a lot of areas in cybersecurity where AI is already providing value. “But how much is it going to cost? We’ve got to work through a lot of details about integrating AI into how we do things,” he says.Too many AI pilots, however, have been deployed without considering those details and defining the use cases. Mogull said he worked with an organization recently that fell into that trap after consultants told management that they could use AI and automation to downsize a particular team by 75%. “Without going into details, the things that team did didn’t have a single use case for AI,” he says. “I couldn’t even imagine a viable use case. It’s one of those areas where generative AI brought zero value.”AI Bubble Soon to Pop?AI companies have continued to feed their models with vast amounts of data and supporting infrastructure to make them more powerful and even achieve artificial general intelligence (AGI). But Marcus, as well as other skeptics, have criticized the “all you need is scale” approach, arguing that not only is it wrong but that it’s created the apparent economic bubble we’re in. And 2026 could be when the spending and investments start to really pull back.”Many people have a huge incentive to keep building the infrastructure, but the vibe has changed,” Marcus says. “Loans will get more expensive, stock prices are coming down, and profits (except for Nvidia) are few and far between.”Even if there’s a bubble, Mogull says AI isn’t a completely overhyped technology and is here to stay. But he remains cautious, given the growth demands on many of the AI companies and risks they pose to the greater economy.”We saw this with the internet and the dot-com boom. Is the Internet here to stay? Hell yes. Are dot coms still around? Yes, they run our lives,” he says. “But it was a rough road getting to the other side if you didn’t play your cards right.”
