Category: English

Source: The Conversation – USA (2) – By Murugan Anandarajan, Professor of Decision Sciences and Management Information Systems, Drexel University

Businesses are acting fast to adopt agentic AI – artificial intelligence systems that work without human guidance – but have been much slower to put governance in place to oversee them, a new survey shows. That mismatch is a major source of risk in AI adoption. In my view, it’s also a business opportunity.

I’m a professor of management information systems at Drexel University’s LeBow College of Business, which recently surveyed more than 500 data professionals through its Center for Applied AI & Business Analytics. We found that 41% of organizations are using agentic AI in their daily operations. These aren’t just pilot projects or one-off tests. They’re part of regular workflows.

At the same time, governance is lagging. Only 27% of organizations say their governance frameworks are mature enough to monitor and manage these systems effectively.

In this context, governance is not about regulation or unnecessary rules. It means having policies and practices that let people clearly influence how autonomous systems work, including who is responsible for decisions, how behavior is checked, and when humans should get involved.

This mismatch can become a problem when autonomous systems act in real situations before anyone can intervene.

For example, during a recent power outage in San Francisco, autonomous robotaxis got stuck at intersections, blocking emergency vehicles and confusing other drivers. The situation showed that even when autonomous systems behave “as designed,” unexpected conditions can lead to undesirable outcomes.

This raises a big question: When something goes wrong with AI, who is responsible – and who can intervene?

Why governance matters

When AI systems act on their own, responsibility no longer lies where organizations expect it. Decisions still happen, but ownership is harder to trace. For instance, in financial services, fraud detection systems increasingly act in real time to block suspicious activity before a human ever reviews the case. Customers often only find out when their card is declined.

So, what if your card is mistakenly declined by an AI system? In that situation, the problem isn’t with the technology itself – it’s working as it was designed – but with accountability. Research on human-AI governance shows that problems happen when organizations don’t clearly define how people and autonomous systems should work together. This lack of clarity makes it hard to know who is responsible and when they should step in.

Without governance designed for autonomy, small issues can quietly snowball. Oversight becomes sporadic and trust weakens, not because systems fail outright, but because people struggle to explain or stand behind what the systems do.

When humans enter the loop too late

In many organizations, humans are technically “in the loop,” but only after autonomous systems have already acted. People tend to get involved once a problem becomes visible – when a price looks wrong, a transaction is flagged or a customer complains. By that point, the system has already been decided, and human review becomes corrective rather than supervisory.

Late intervention can limit the fallout from individual decisions, but it rarely clarifies who is accountable. Outcomes may be corrected, yet responsibility remains unclear.

Recent guidance shows that when authority is unclear, human oversight becomes informal and inconsistent. The problem is not human involvement, but timing. Without governance designed upfront, people act as a safety valve rather than as accountable decision-makers.

How governance determines who moves ahead

Agentic AI often brings fast, early results, especially when tasks are first automated. Our survey found that many companies see these early benefits. But as autonomous systems grow, organizations often add manual checks and approval steps to manage risk.

Over time, what was once simple slowly becomes more complicated. Decision-making slows down, work-arounds increase, and the benefits of automation fade. This happens not because the technology stops working, but because people never fully trust autonomous systems.

This slowdown doesn’t have to happen. Our survey shows a clear difference: Many organizations see early gains from autonomous AI, but those with stronger governance are much more likely to turn those gains into long-term results, such as greater efficiency and revenue growth. The key difference isn’t ambition or technical skills, but being prepared.

Good governance does not limit autonomy. It makes it workable by clarifying who owns decisions, how systems function is monitored, and when people should intervene. International guidance from the OECD – the Organization for Economic Cooperation and Development – emphasizes this point: Accountability and human oversight need to be designed into AI systems from the start, not added later.

Rather than slowing innovation, governance creates the confidence organizations need to extend autonomy instead of quietly pulling it back.

The next advantage is smarter governance

The next competitive advantage in AI will not come from faster adoption, but from smarter governance. As autonomous systems take on more responsibility, success will belong to organizations that clearly define ownership, oversight and intervention from the start.

In the era of agentic AI, confidence will accrue to the organizations that govern best, not simply those that adopt first.

Murugan Anandarajan does not work for, consult, own shares in or receive funding from any company or organization that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.

– ref. Companies are already using agentic AI to make decisions, but governance is lagging behind – https://theconversation.com/companies-are-already-using-agentic-ai-to-make-decisions-but-governance-is-lagging-behind-272792

Source: The Conversation – USA (2) – By Ahmed Elgammal, Professor of Computer Science and Director of the Art & AI Lab, Rutgers University

Generative AI was trained on centuries of art and writing produced by humans.

But scientists and critics have wondered what would happen once AI became widely adopted and started training on its outputs.

A new study points to some answers.

In January 2026, artificial intelligence researchers Arend Hintze, Frida Proschinger Åström and Jory Schossau published a study showing what happens when generative AI systems are allowed to run autonomously – generating and interpreting their own outputs without human intervention.

The researchers linked a text-to-image system with an image-to-text system and let them iterate – image, caption, image, caption – over and over and over.

Regardless of how diverse the starting prompts were – and regardless of how much randomness the systems were allowed – the outputs quickly converged onto a narrow set of generic, familiar visual themes: atmospheric cityscapes, grandiose buildings and pastoral landscapes. Even more striking, the system quickly “forgot” its starting prompt.

The researchers called the outcomes “visual elevator music” – pleasant and polished, yet devoid of any real meaning.

For example, they started with the image prompt, “The Prime Minister pored over strategy documents, trying to sell the public on a fragile peace deal while juggling the weight of his job amidst impending military action.” The resulting image was then captioned by AI. This caption was used as a prompt to generate the next image.

After repeating this loop, the researchers ended up with a bland image of a formal interior space – no people, no drama, no real sense of time and place.

As a computer scientist who studies generative models and creativity, I see the findings from this study as an important piece of the debate over whether AI will lead to cultural stagnation.

The results show that generative AI systems themselves tend toward homogenization when used autonomously and repeatedly. They even suggest that AI systems are currently operating in this way by default.

The familiar is the default

This experiment may appear beside the point: Most people don’t ask AI systems to endlessly describe and regenerate their own images. The convergence to a set of bland, stock images happened without retraining. No new data was added. Nothing was learned. The collapse emerged purely from repeated use.

But I think the setup of the experiment can be thought of as a diagnostic tool. It reveals what generative systems preserve when no one intervenes.

This has broader implications, because modern culture is increasingly influenced by exactly these kinds of pipelines. Images are summarized into text. Text is turned into images. Content is ranked, filtered and regenerated as it moves between words, images and videos. New articles on the web are now more likely to be written by AI than humans. Even when humans remain in the loop, they are often choosing from AI-generated options rather than starting from scratch.

The findings of this recent study show that the default behavior of these systems is to compress meaning toward what is most familiar, recognizable and easy to regenerate.

Cultural stagnation or acceleration?

For the past few years, skeptics have warned that generative AI could lead to cultural stagnation by flooding the web with synthetic content that future AI systems then train on. Over time, the argument goes, this recursive loop would narrow diversity and innovation.

Champions of the technology have pushed back, pointing out that fears of cultural decline accompany every new technology. Humans, they argue, will always be the final arbiter of creative decisions.

What has been missing from this debate is empirical evidence showing where homogenization actually begins.

The new study does not test retraining on AI-generated data. Instead, it shows something more fundamental: Homogenization happens before retraining even enters the picture. The content that generative AI systems naturally produce – when used autonomously and repeatedly – is already compressed and generic.

This reframes the stagnation argument. The risk is not only that future models might train on AI-generated content, but that AI-mediated culture is already being filtered in ways that favor the familiar, the describable and the conventional.

Retraining would amplify this effect. But it is not its source.

This is no moral panic

Skeptics are right about one thing: Culture has always adapted to new technologies. Photography did not kill painting. Film did not kill theater. Digital tools have enabled new forms of expression.

But those earlier technologies never forced culture to be endlessly reshaped across various mediums at a global scale. They did not summarize, regenerate and rank cultural products – news stories, songs, memes, academic papers, photographs or social media posts – millions of times per day, guided by the same built-in assumptions about what is “typical.”

The study shows that when meaning is forced through such pipelines repeatedly, diversity collapses not because of bad intentions, malicious design or corporate negligence, but because only certain kinds of meaning survive the text-to-image-to-text repeated conversions.

This does not mean cultural stagnation is inevitable. Human creativity is resilient. Institutions, subcultures and artists have always found ways to resist homogenization. But in my view, the findings of the study show that stagnation is a real risk – not a speculative fear – if generative systems are left to operate in their current iteration.

They also help clarify a common misconception about AI creativity: Producing endless variations is not the same as producing innovation. A system can generate millions of images while exploring only a tiny corner of cultural space.

In my own research on creative AI, I found that novelty requires designing AI systems with incentives to deviate from the norms. Without it, systems optimize for familiarity because familiarity is what they have learned best. The study reinforces this point empirically. Autonomy alone does not guarantee exploration. In some cases, it accelerates convergence.

This pattern already emerged in the real world: One study found that AI-generated lesson plans featured the same drift toward conventional, uninspiring content, underscoring that AI systems converge toward what’s typical rather than what’s unique or creative.

Lost in translation

Whenever you write a caption for an image, details will be lost. Likewise for generating an image from text. And this happens whether it’s being performed by a human or a machine.

In that sense, the convergence that took place is not a failure that’s unique to AI. It reflects a deeper property of bouncing from one medium to another. When meaning passes repeatedly through two different formats, only the most stable elements persist.

But by highlighting what survives during repeated translations between text and images, the authors are able to show that meaning is processed inside generative systems with a quiet pull toward the generic.

The implication is sobering: Even with human guidance – whether that means writing prompts, selecting outputs or refining results – these systems are still stripping away some details and amplifying others in ways that are oriented toward what’s “average.”

If generative AI is to enrich culture rather than flatten it, I think systems need to be designed in ways that resist convergence toward statistically average outputs. There can be rewards for deviation and support for less common and less mainstream forms of expression.

The study makes one thing clear: Absent these interventions, generative AI will continue to drift toward mediocre and uninspired content.

Cultural stagnation is no longer speculation. It’s already happening.

Ahmed Elgammal does not work for, consult, own shares in or receive funding from any company or organization that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.

– ref. AI-induced cultural stagnation is no longer speculation − it’s already happening – https://theconversation.com/ai-induced-cultural-stagnation-is-no-longer-speculation-its-already-happening-272488

Source: The Conversation – USA – By Saman Zonouz, Associate Professor of Cybersecurity and Privacy and Electrical and Computer Engineering, Georgia Institute of Technology

The darkness that swept over the Venezuelan capital in the predawn hours of Jan. 3, 2026, signaled a profound shift in the nature of modern conflict: the convergence of physical and cyber warfare. While U.S. special operations forces carried out the dramatic seizure of Venezuelan President Nicolás Maduro, a far quieter but equally devastating offensive was taking place in the unseen digital networks that help operate Caracas.

The blackout was not the result of bombed transmission towers or severed power lines but rather a precise and invisible manipulation of the industrial control systems that manage the flow of electricity. This synchronization of traditional military action with advanced cyber warfare represents a new chapter in international conflict, one where lines of computer code that manipulate critical infrastructure are among the most potent weapons.

To understand how a nation can turn an adversary’s lights out without firing a shot, you have to look inside the controllers that regulate modern infrastructure. They are the digital brains responsible for opening valves, spinning turbines and routing power.

For decades, controller devices were considered simple and isolated. Grid modernization, however, has transformed them into sophisticated internet-connected computers. As a cybersecurity researcher, I track how advanced cyber forces exploit this modernization by using digital techniques to control the machinery’s physical behavior.

Hijacked machines

My colleagues and I have demonstrated how malware can compromise a controller to create a split reality. The malware intercepts legitimate commands sent by grid operators and replaces them with malicious instructions designed to destabilize the system.

For example, malware could send commands to rapidly open and close circuit breakers, a technique known as flapping. This action can physically damage massive transformers or generators by causing them to overheat or go out of sync with the grid. These actions can cause fires or explosions that take months to repair.

Simultaneously, the malware calculates what the sensor readings should look like if the grid were operating normally and feeds these fabricated values back to the control room. The operators likely see green lights and stable voltage readings on their screens even as transformers are overloading and breakers are tripping in the physical world. This decoupling of the digital image from physical reality leaves defenders blind, unable to diagnose or respond to the failure until it is too late.

Historical examples of this kind of attack include the Stuxnet malware that targeted Iranian nuclear enrichment plants. The malware destroyed centrifuges in 2009 by causing them to spin at dangerous speeds while feeding false “normal” data to operators.

Another example is the Industroyer attack by Russia against Ukraine’s energy sector in 2016. Industroyer malware targeted Ukraine’s power grid, using the grid’s own industrial communication protocols to directly open circuit breakers and cut power to Kyiv.

More recently, the Volt Typhoon attack by China against the United States’ critical infrastructure, exposed in 2023, was a campaign focused on pre-positioning. Unlike traditional sabotage, these hackers infiltrated networks to remain dormant and undetected, gaining the ability to disrupt the United States’ communications and power systems during a future crisis.

To defend against these types of attacks, the U.S. military’s Cyber Command has adopted a “defend forward” strategy, actively hunting for threats in foreign networks before they reach U.S. soil.

Domestically, the Cybersecurity and Infrastructure Security Agency promotes “secure by design” principles, urging manufacturers to eliminate default passwords and utilities to implement “zero trust” architectures that assume networks are already compromised.

Supply chain vulnerability

Nowadays, there is a vulnerability lurking within the supply chain of the controllers themselves. A dissection of firmware from major international vendors reveals a significant reliance on third-party software components to support modern features such as encryption and cloud connectivity.

This modernization comes at a cost. Many of these critical devices run on outdated software libraries, some of which are years past their end-of-life support, meaning they’re no longer supported by the manufacturer. This creates a shared fragility across the industry. A vulnerability in a single, ubiquitous library like OpenSSL – an open-source software toolkit used worldwide by nearly every web server and connected device to encrypt communications – can expose controllers from multiple manufacturers to the same method of attack.

Modern controllers have become web-enabled devices that often host their own administrative websites. These embedded web servers present an often overlooked point of entry for adversaries.

Attackers can infect the web application of a controller, allowing the malware to execute within the web browser of any engineer or operator who logs in to manage the plant. This execution enables malicious code to piggyback on legitimate user sessions, bypassing firewalls and issuing commands to the physical machinery without requiring the device’s password to be cracked.

The scale of this vulnerability is vast, and the potential for damage extends far beyond the power grid, including transportation, manufacturing and water treatment systems.

Using automated scanning tools, my colleagues and I have discovered that the number of industrial controllers exposed to the public internet is significantly higher than industry estimates suggest. Thousands of critical devices, from hospital equipment to substation relays, are visible to anyone with the right search criteria. This exposure provides a rich hunting ground for adversaries to conduct reconnaissance and identify vulnerable targets that serve as entry points into deeper, more protected networks.

The success of recent U.S. cyber operations forces a difficult conversation about the vulnerability of the United States. The uncomfortable truth is that the American power grid relies on the same technologies, protocols and supply chains as the systems compromised abroad.

Regulatory misalignment

The domestic risk, however, is compounded by regulatory frameworks that struggle to address the realities of the grid. A comprehensive investigation into the U.S. electric power sector my colleagues and I conducted revealed significant misalignment between compliance with regulations and actual security. Our study found that while regulations establish a baseline, they often foster a checklist mentality. Utilities are burdened with excessive documentation requirements that divert resources away from effective security measures.

This regulatory lag is particularly concerning given the rapid evolution of the technologies that connect customers to the power grid. The widespread adoption of distributed energy resources, such as residential solar inverters, has created a large, decentralized vulnerability that current regulations barely touch.

Analysis supported by the Department of Energy has shown that these devices are often insecure. By compromising a relatively small percentage of these inverters, my colleagues and I found that an attacker could manipulate their power output to cause severe instabilities across the distribution network. Unlike centralized power plants protected by guards and security systems, these devices sit in private homes and businesses.

Accounting for the physical

Defending American infrastructure requires moving beyond the compliance checklists that currently dominate the industry. Defense strategies now require a level of sophistication that matches the attacks. This implies a fundamental shift toward security measures that take into account how attackers could manipulate physical machinery.

The integration of internet-connected computers into power grids, factories and transportation networks is creating a world where the line between code and physical destruction is irrevocably blurred.

Ensuring the resilience of critical infrastructure requires accepting this new reality and building defenses that verify every component, rather than unquestioningly trusting the software and hardware – or the green lights on a control panel.

Saman Zonouz receives funding from the Department of Energy Office of Cybersecurity, Energy Security, and Emergency Response (DOE CESER) and the National Science Foundation (NSF).

– ref. Hacking the grid: How digital sabotage turns infrastructure into a weapon – https://theconversation.com/hacking-the-grid-how-digital-sabotage-turns-infrastructure-into-a-weapon-272874

Source: The Conversation – USA – By Micah Altman, Research Scientist, MIT Libraries, Massachusetts Institute of Technology (MIT)

A growing distrust of expertise is reshaping the terrain of science in the United States.

Since the pandemic, the partisan divide over science has widened dramatically. While 77% of Americans have at least a fair amount of confidence that scientists act in the best interests of the public, that breaks down to 90% of Democrats and 65% of Republicans.

If people think scientists are trying to impose their political beliefs rather than expressing honest scientific judgments in the pursuit of truth, public trust in expert consensus will continue to erode.

With recent events, such as Robert F. Kennedy Jr. replacing the expert vaccine panel at the Department of Health and Human Services, and the Trump administration threatening to withdraw research funding from universities that don’t follow its ideological dictates, the political divide in public perception may grow even deeper.

As social scientists who study the role of science in society, we are deeply concerned about the decline of public trust in expertise, which is often fueled by politicians who manipulate people’s suspicions about experts. Skepticism is sometimes justified, of course. But a system based on expertise is the best one modern democracies have come up with to offer guidance on the various complex issues they face.

Who is an expert?

Before you can place your trust in a community of experts, you need a way to determine who counts as an expert. Modern societies usually do this through a sequence of training within accredited schools and universities – institutions whose reputations depend on their ability to train reliable and trustworthy experts.

Unlike the ancient alchemists’ guilds or modern elites, science is not secret, nor gated by family descent or social ties. Today anyone is permitted to become a scientific expert by attaining academic degrees and certifications and establishing a public track record of published research, teaching and contributions to one’s field.

The government also plays a critical role by requiring doctors or engineers to hold certain degrees or by granting universities formal quality certifications, such as accreditation. As an individual, you can’t evaluate the trustworthiness of every person claiming to have expertise – whether a heart surgeon or an electrician. The governmental license carried by these professionals makes that unnecessary.

In any field of knowledge, there is a web of legitimacy, knotted together by visible signals of trust, such as degrees, publications, affiliations and accreditations. Expertise is a team sport.

What is expert consensus?

The most reliable guidance is based on a rigorous group decision-making process, in which people with diverse training and experience contribute their expertise to a dialogue aimed at reaching consensus. The scientific approach to consensus is transparent and deliberate: Scientific consensus processes – such as the National Academies consensus study process, or a PRISMA review – are systematic in incorporating the credible evidence that is available and synthesizing different expert judgments.

The system, honed over decades, is based on the theory that better decisions can be achieved by systematically aggregating many independent opinions – if the group is well trained, draws from a common body of evidence, relies on a common understanding of research practices, and each of its members are able to independently weigh the evidence.

Such communities of experts arise in many settings, from engineers recommending building codes to epidemiologists proposing policies to contain a viral epidemic.

An expert community doesn’t need everyone to be right – or even to agree – in every case for the process to generate useful results. As long as each person is usually right and the community deliberates systematically on the basis of reason and evidence, the resulting consensus will be the best that can be achieved within the limits of current knowledge.

In short, expert consensus requires trained experts, common evidence and systematic deliberation.

Professional consensus vs. individual opinion

Expert consensus doesn’t mean that experts agree on everything, or that everyone must agree with the experts. In a democracy, expert advice is valuable, but it’s not the last word.

The U.S. Bill of Rights enshrines the idea that freedom of speech is fundamental to good government and to leading good lives. But there’s a distinction between speaking one’s mind and speaking from authority. Experts have a right to express their personal opinions and also a duty to exercise care when speaking in areas of their expertise.

This distinction is at issue in the Chiles vs Salazar case before the Supreme Court. It centers on a Colorado state law that prohibits so-called “conversion therapy” for gay or trans children.

Does doing so violate the free speech rights of therapists? It’s not illegal to believe trans children can be talked out of being trans, it’s just illegal to pursue that practice as a licensed professional, because medical experts have reached a consensus that conversion therapy is both useless and harmful.

Expert consensus is necessary to make sound decisions based on science and evidence, but that doesn’t mean experts must abstain from politics or refrain from expressing dissenting opinions. In fact, political restrictions on scientific debate weaken science, as seen in repressive societies.

What does expert consensus provide?

In our fractious political climate, people sometimes think divergent expert opinions mean that consensus does not exist, or no experts can be trusted. Some people say, “Do your own research,” which often leads to rejecting consensus and falling victim to conspiracies and disinformation.

In practice, consensus is compatible with substantial disagreement. In many fields, scientific consensus deals with broad patterns rather than individual cases. For example, medical experts may agree on the nature of a specific condition, and the average efficacy of a given treatment, yet make different predictions about the benefits for a specific patient.

Society faces pressing questions about the behavior of complex and uncertain systems: How much is climate likely to change if CO₂ emissions continue at the current rate – and what ecological changes should we expect? What accounts for changing cancer rates – and what are the most promising paths to develop a broad “cure”? Are AIs developing intelligence and self-awareness – and how can they be designed to be behave safely? What social institutions are essential for human flourishing – and how can they be preserved?

It’s the fundamental role of democratic government to determine which goals we as a society pursue and how to balance competing values. And when we face high-stakes issues involving complex systems and uncertain approaches, scientific expert consensus can act as an honest broker to provide a menu of possible approaches and predictions for each one’s likely consequences.

Micah Altman received research funding from the National Science Foundation and the Andrew W Mellon Foundation to conduct research related to the science of science, and related to open science.

Philip N. Cohen does not work for, consult, own shares in or receive funding from any company or organization that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.

– ref. ‘Expertise’ shouldn’t be a bad word – expert consensus guides science and society – https://theconversation.com/expertise-shouldnt-be-a-bad-word-expert-consensus-guides-science-and-society-271467