Technology

US law students returning to campus this autumn have encountered sharply different approaches to artificial intelligence, with some schools restricting laptops and phones in class while others require students to use AI in certain courses, Reuters reported.

Some students are facing both approaches simultaneously as law schools seek to balance traditional legal education with training in a technology that is rapidly transforming the legal profession.

At least a dozen US law schools introduced new or revised AI policies over the summer, setting out when and how students may use the technology in their studies or requiring them to take courses covering AI.

The moves come amid a wider push across higher education to respond to AI, including the opportunities it presents and the risks it poses to student learning. A report on AI and education from the Massachusetts Institute of Technology last month said the ease of obtaining answers from chatbots can induce "cognitive surrender".

“It certainly is a time of fundamental questioning of what we do and how we do it,” said University of Georgia law dean Usha Rodrigues, whose school released a new AI policy in August that includes a default ban on devices in the classroom and requires students to take at least one AI-related technology course.

The “analog classroom” policy was partly prompted by stories of students consulting AI during the traditional law school practice of professors cold-calling students in class, Rodrigues said.

“It's not really about getting the right answer,” she said of the Socratic method long used in law schools. “It’s about being on the spot, thinking on your feet, sometimes getting it right, sometimes getting it wrong, and listening to your colleagues and their answers and evaluating them critically.”

The stakes are especially high for law schools because lawyers have professional and ethical obligations relating to their competence. They may defend clients in death penalty cases or represent corporations in billion-dollar litigation or deals, while confidentiality rules also affect how they can use AI.

The University of California, Berkeley School of Law in May unveiled a particularly restrictive AI policy that bars students from using AI to brainstorm a paper topic, summarise a legal rule for use in a paper, correct grammatical mistakes or identify repetitive passages.

The law school also offers a range of AI courses, but its student-run campus club, AI at Berkeley Law, said in a statement to Reuters that the default rules run counter to the school's positioning as a leader in technology and law, and make some students "recoil".

“There’s a difference between outsourcing legal reasoning to AI and using AI as a tool within a legal workflow, and legal education should teach that distinction rather than avoid it,” the group said.

Berkeley Law professor Colleen Chien, who serves on the school's AI Leadership Committee, told Reuters that faculty members are free to deviate from the default policy if doing so serves their teaching and learning goals. "The policy is the starting point of a conversation, it's not the end of it," Chien said.

Similar AI clubs have emerged at other law schools, reflecting students' concerns that the technology will be necessary for their careers as law firms race to incorporate AI into legal practice. Law firms and other employers now expect law school graduates to have a basic familiarity with AI, University of Texas law dean Bobby Chesney said during a recent webinar on AI in law schools.

A New Reality

Suffolk University Law School dean Andrew Perlman used ChatGPT to compile publicly available law school AI policies and in August released an online database containing information from 180 law schools.

A growing number of law schools now require students to receive instruction on AI during orientation, in mandatory first-year courses or through standalone AI classes. At least 36 have some form of mandated AI instruction, according to the database.

“I think what's happened over the last few years is an acceptance and an expectation that AI is here to stay,” Perlman said.

The University of Chicago Law School in July became the first to ban laptops and phones from required first-year classes to ensure that “students actually learn to think critically, strategically, and independently without relying on AI”, before teaching them to use those tools later on, according to the school's new policy.

Columbia Law School’s new AI policy, released in August, allows students to use AI to “test arguments, solicit criticism, and explore alternative articulations” as well as correct spelling and grammar in their written work. But that work must be “student generated” and reflect “ideas, arguments, analysis, and expression that are the product of the student's own intellectual creativity and judgment.”

The University of Michigan takes a middle ground by allowing students to use AI to brainstorm and research papers but not to draft, edit or revise their work.

Avyay Casheekar, president of Michigan Law's AI Law and Policy Society, said the school's policy is a "safe position" that appears to be in line with those of other law schools. But he worried both that over-reliance on AI could stunt learning and that excessive caution could stymie "inventive" uses of the technology.

Nearly all schools ban AI use in exams, said University of Houston law professor Seth Chandler, who writes a blog about AI in legal education. Once-common take-home exams have given way to in-class tests using software that blocks internet access on students’ laptops, preventing AI use, he said.

“Faculty do not want students taking exams with AI in front of them, except maybe in some unusual cases,” Chandler said. “The legal material needs to be in your brain at the time that you are assessed.”

Growing concern in the US over unchecked and rapidly developing artificial intelligence models is increasing the urgency for companies using AI to put their houses in order, according to a Bloomberg Law report.

Warnings from leaders of top AI labs about existential risks posed by the most advanced models mark an inflection point for companies facing competitive pressure to accelerate AI adoption while aggressively re-examining their safety guardrails.

Legal counsel are advising companies to mitigate their risks by following core practices: use pioneering AI models from leaders such as Anthropic and OpenAI cautiously, keep a human in the loop, vet vendors and carefully review the level of access to company information given to AI agents. And, they say, companies should not panic.

“A lot of people are worried that the sky is falling, and it may be, but our job is to make sure the roof doesn’t cave in,” said Eric Dodson Greenberg, executive vice president and chief legal officer of Cox Media Group.

Anthropic CEO Dario Amodei, who published an essay on Saturday urging a slowdown in AI growth, has acknowledged the challenge facing companies. He and Salesforce CEO Marc Benioff said on Tuesday that companies have only just begun taking advantage of AI and could need more help to make that transformation.

Still, at this point, there is no turning back for companies using AI in pursuit of efficiency gains to remain competitive and meet the expectations of their boards and executive teams.

“The toothpaste is out of the tube when it comes to AI deployment across companies,” said Virginia Johnson, a partner at OGC, a firm that includes general counsel providing legal help to companies. “That means for the GC, the primary mandate is having proper governance internally to make sure that your use of AI is done according to policies, procedures, proper training, ethical deployment, and managing the risk within your own company.”

Uneven Scenario

A thorough examination of how companies are keeping AI in check cannot come soon enough. An EY survey released recently pointed to an uneven picture as companies try to balance speed and safety. About half of AI leaders at companies said their organisations had sidestepped AI governance frameworks to deploy AI quickly.

It is another warning sign for companies navigating the rapid evolution of AI. “We have to be embracing it through incremental calibrated steps and safe experiments that allow us to move this forward where we’re managing the risk level,” said Cox Media’s Greenberg, who is also a Bloomberg Law columnist.

His advice is to think of AI governance as an ongoing process because the technology is constantly changing, as is the way humans interact with it.

That means examining exactly where AI is being used within a company, identifying the highest-risk uses, assigning a human owner, vetting vendors and establishing a process for identifying problems, said Jobe Danganan, a former founding enforcement attorney at the Consumer Financial Protection Bureau who is now co-founder and CEO of LexText AI, a legal text platform.

“You won’t eliminate every risk, but you should understand the risks you’re taking and be able to explain why your safeguards were reasonable,” said Danganan, who previously served as general counsel for financial technology companies.

Slowing down could mean that the next frontier model is delayed, potentially leading to higher costs. “There’s going to be more cost in the system that maybe we have to absorb as a customer because OpenAI, Anthropic, etc., are going to have to absorb more cost for governance, pass it along to us possibly,” said Dana Rao, former general counsel and executive vice president of Adobe.

“We have to have a hard conversation now about the earnings we just reported, the forecast we just reported. Are they still good? We have to go back to the street. We have to be transparent to our investors,” said Rao, an AI fellow at Fordham University.

Still, many companies using standard AI models will be fine, legal leaders said.

“Generally speaking, if I’ve successfully deployed today’s version of an LLM, I would feel pretty comfortable saying just keep using it,” Rao said.

Shared Responsibility

The public nature of the disclosures last weekend could help companies talk more openly about AI risks, said Scott Meyers, chairman and CEO of Akerman LLP.

“What it means is that we all need to work together because this goes candidly beyond just helping a company survive or meet its revenue targets,” he said. “The risk here goes far beyond just how it impacts the company, but all of the companies now are aware of what this can mean, both from an external attack plus using agents inside the firewall.”

Political leaders have so far differed over when and how Congress should intervene with increased regulation. President Donald Trump has argued against regulation, saying it would help China.

Separately, technology leaders such as Demis Hassabis of Google DeepMind have advocated for a regulatory body modelled on Finra, which monitors US brokerage firms and stockbrokers.

“It’s very much in the US government’s interest to be at the forefront of this conversation and the convener and the drafter. And if they don’t, I do think the industry is going to have to look elsewhere to help them come up with industry standards,” said Beth George, head of US litigation, arbitration and global investigations at Freshfields.

A federal judiciary task force has made dozens of recommendations about courts’ use of artificial intelligence, the group’s leader said.

Judge Sidney Thomas of the Ninth Circuit, who chairs the judiciary’s AI task force, said the group may issue final guidance by as soon as the end of this year on what courts might want to do to address use of the platforms.

He said interim guidance that the group issued last year addresses judges’ use of the tools, by making it clear that the platforms can’t be used for “core functions” like deciding cases. But Thomas said AI seems to be most effective in handling administrative tasks in the courts.

“Publicly, I think people have focused on the use in chambers, but actually I think the more important and significant use may be in court operations,” Thomas said during a federal judiciary press briefing Thursday.

Some court clerks last year have experimented using AI for their administrative tasks, Bloomberg Law reported. Some courts have publicly rolled out use of AI for some functions: The Fifth Circuit, for example, offers the use of AI to help attorneys properly file documents.

The federal judiciary described the AI task force’s interim guidance last year after a pair of judges acknowledged that the tools were used by their staff, resulting in fake or incorrect case citations.

Thursday’s press briefing took place after the biannual meeting of the Judicial Conference, the courts’ policy-making body. That group heard a presentation about the AI task force at its meeting, held at the US Supreme Court.

Thomas told reporters that the task force is solely an advisory body, and its recommendations will have to be considered by other judiciary committees in order to be formally implemented.

Courts have also been facing AI-generated content from attorneys and self-represented litigants, with judges issuing discipline in some cases.

Chief Judge Jeffrey Sutton, who chairs the Judicial Conference’s Executive Committee, said AI use by litigants is an issue that may be addressed by potential changes to the courts’ procedural rules.

The Administrative Office of the US Courts also said it will start rolling out this year a new case management system for court documents, after repeated hacks targeted the current infrastructure for hosting filings. Bloomberg Law reported last year that hackers targeted sensitive, sealed court documents, using the same vulnerabilities that were exploited in a previous breach.

All new district court cases will be in the new case management system by the end of next year, according to the judiciary’s press release, and appellate and bankruptcy courts will then adopt it.

Judge Robert J. Conrad, director of the AO, said at the press briefing that the changes will take place on the back-end of the courts’ websites, and users will see “little to no difference in the way that system is accessed.”

“We can offer litigants, cooperating witnesses in criminal cases, businesses with trade secrets, the confidence they deserve with their case is tried in federal court,” Conrad said.

A federal judge has ordered Google to relax rules governing its online advertising auctions and appoint an internal antitrust compliance monitor, while stopping short of requiring the Alphabet unit to break up its advertising technology business.

US District Judge Leonie Brinkema in Alexandria, Virginia, set out the remedies in a 106-page decision unsealed on Wednesday, two weeks after rejecting the US Department of Justice's demand that Google break up its advertising technology business.

Brinkema said Google should instead change some of its business practices, after finding in April 2025 that the company maintained an illegal monopoly over parts of the online advertising technology market.

The remedies "will be sufficient to effectively pry open to competition the ad tech markets that were injured by Google's unlawful conduct, and prevent Google from reverting to anticompetitive conduct in these markets", Brinkema wrote.

Second Judge Rejects Breakup

Google said on Wednesday that it disagreed with Brinkema's liability ruling concerning its Google Ad Manager publishing tool and would appeal. The company also maintained that forcing a divestiture would have made it harder for small businesses to reach customers.

Associate Attorney General Stanley Woodward Jr said in a statement that the decision was a "significant victory" in the Justice Department's efforts to protect and restore competition. He added that the department was reviewing the opinion to determine its legal options.

The decision spared Google from having to break up another part of its internet business as the Mountain View, California-based company races to expand in artificial intelligence against rivals including Anthropic and OpenAI.

Last September, a different judge ordered Google to open up competition in online search, but declined to require the sale of its widely used Chrome browser.

Annual global digital advertising spending could grow to $605 billion next year from $424 billion in 2023, according to Brinkema's decision.

Advertising accounted for about 73% of Alphabet's revenue last year. The company's market value exceeds $4.1 trillion.

Six Years

The government had sought to force Google to sell AdX, where publishers pay a 20% fee to sell ads through auctions that take place instantly when users load websites. The government argued that Google could not be trusted to operate the service.

Brinkema rejected that remedy, saying that allowing other publisher ad servers to access real-time bids from AdX would restore "much-needed" competition.

The judge accepted proposals requiring Google not to force websites that use its ad server to also use AdX. Google would also have to end practices that publishers had complained kept them locked into its advertising technology tools.

Brinkema also said an internal compliance monitor was necessary given the "gravity" of Google's antitrust violations, although the monitor would have less oversight than the government had sought.

The changes must remain in place for six years, rather than the 15 years sought by the Justice Department and several states that also sued Google.

After issuing her ruling, Brinkema gave both sides 14 days to seek redactions of confidential information from the written decision and 30 days to file a proposed final judgment reflecting the remedies she has ordered.

As artificial intelligence grows more powerful, researchers have documented cases in which AI models have attempted to deceive users, evade restrictions on their use or access other computer systems. The question is whether companies are required under US law to tell the public or regulators when such events occur.

No single federal law is aimed specifically at companies such as Anthropic or OpenAI, which are developing highly capable AI systems. There is also no broad US legal requirement for AI developers to publicly disclose dangerous model behaviour, alarming new capabilities, deceptive conduct or other activities if they have not already resulted in concrete harm.

Federal legislation has been introduced that would require AI companies to report dangerous behaviour, such as attempts to evade human oversight. The bill's sponsor described it as a "catch-it-early and sound-the-alarm bill". However, there is currently no general incident-reporting system requiring companies to disclose dangerous AI behaviour when it is discovered.

Lawmakers have been debating stronger controls since July, when OpenAI said rogue AI agents had bypassed internal controls, reached the open internet and compromised the infrastructure of AI startup Hugging Face. Outside researchers have since identified additional incidents alleged to involve OpenAI-linked agents, while Anthropic has reported that some of its Claude models hacked into the systems of three companies during cybersecurity tests.

When Would An AI Incident Trigger Mandatory Disclosure?

Legal frameworks that already apply generally to US companies can govern certain types of AI-related incidents. Under US Securities and Exchange Commission rules, public companies must disclose cybersecurity incidents within four business days if they determine that an incident is material to investors. The disclosure must cover the nature, scope and timing of the incident, as well as its likely impact on the company, its financial condition and results of operations.

Some US states have also begun regulating AI companies. A new California law requires AI companies with more than $500 million in revenue to disclose how they assess the risks that their technology could escape human control or aid the development of bioweapons, and to make those assessments available to the public. The law allows fines of up to $1 million per violation.

What If Private Data Is Exposed?

All 50 US states have laws requiring companies to notify individuals, and in some cases regulators, about data security breaches that expose certain types of personal information. The requirements vary by state, and there is no comprehensive federal data-breach notification requirement.

Federal statutes also require certain companies in sectors such as healthcare and finance to notify individuals or regulators when personal information is compromised. Those reporting requirements can apply to AI companies themselves or to any other company that experiences a qualifying breach.

What Other Regulators Could Take Action?

The US Federal Trade Commission, which enforces consumer-protection laws, has authority to pursue companies over unfair or deceptive practices. That authority could apply if a company is suspected of misrepresenting the safety of its AI systems by concealing known security weaknesses or other dangers, or by making claims about safeguards that prove inaccurate.

If an alleged crime were committed by an autonomous AI system, the US Justice Department could use existing fraud, securities and cyber-enforcement statutes. Prosecutors could argue that the AI company responsible for creating the system knowingly or recklessly allowed the misconduct to occur.

What Gaps Remain In Existing Disclosure Rules?

A company that discovers alarming AI behaviour during testing may have no clear obligation to disclose it publicly if there is no data breach, investor impact, consumer harm or sector-specific reporting trigger.

US Senate lawmakers are considering legislation that would require AI companies to demonstrate that they have taken reasonable steps to prevent their systems from causing harm. One proposal would empower the Secretary of Commerce to seek evidence that AI companies are taking precautions to prevent harm under a "duty of care" standard.

A federal judge in Texas has ordered Elon Musk’s X Corp and SpaceXAI to disclose an agreement they reached with Apple to resolve antitrust claims against the iPhone maker, after Apple’s co-defendant OpenAI sought access to the deal to strengthen its defence in the case.

US District Judge Mark Pittman in Fort Worth said X must file with the court any agreement or combination of agreements with Apple relating to the resolution of the plaintiffs’ claims against the company.

X this week resolved its lawsuit against Apple without revealing the terms and said it would continue pursuing claims against OpenAI in the case.

The lawsuit, filed last year, alleged that Apple violated antitrust law by exclusively integrating OpenAI’s ChatGPT into Apple Intelligence features on iPhones and other Apple devices.

In a court filing on Tuesday, OpenAI’s lawyers at Wachtell asked Pittman to order disclosure of any settlement agreement.

OpenAI said it wants to see the terms to help rebut Musk’s pending claims against the company. It also said X’s apparent agreement with Apple could undermine any effort by X to seek a monetary payment from OpenAI. X and SpaceXAI were ordered to respond to OpenAI’s request by Thursday.

Companies Deny Wrongdoing

Apple and OpenAI have each denied wrongdoing in the lawsuit. Apple has previously said its agreement with OpenAI was not exclusive.

In May, OpenAI defeated a separate lawsuit brought by Musk that accused the company of straying from its original mission of developing artificial intelligence for the benefit of humanity rather than for profit.

Apple is also suing OpenAI in a separate case over alleged trade secret theft. OpenAI has denied those claims.

Artificial intelligence is rapidly changing the modern workplace. From screening thousands of job applications to analysing employee performance, predicting attrition and even recommending disciplinary action, AI is increasingly being placed in positions once reserved for human managers and HR professionals.

The attraction is obvious: AI can process enormous quantities of information quickly, identify patterns and potentially reduce human inconsistency. But when an algorithm determines whether an individual should be hired, promoted, disciplined or dismissed, a fundamental question arises: can an employer allow AI to make decisions about employees without meaningful human review?

The answer is becoming increasingly complex. As AI moves deeper into employment decision-making, organisations must balance technological efficiency with fairness, transparency, privacy and accountability.

AI Recruitment: Efficient, But Not Necessarily Neutral

AI recruitment systems can screen CVs, rank candidates, analyse applications and identify individuals who appear suitable for a particular role. For employers handling large volumes of applications, this can significantly reduce recruitment time and administrative costs.

However, an algorithm is only as objective as the data and assumptions behind it. If an AI system is trained on historical hiring data reflecting existing workplace inequalities, it may reproduce or even amplify those patterns. A system could unintentionally disadvantage candidates based on factors such as gender, age, disability, nationality, educational background or career history.

This creates a difficult legal question: who is responsible when an algorithm discriminates? The employer cannot simply argue that “the AI made the decision”. Organisations remain responsible for the systems they deploy and the employment decisions made through them.

Performance Scoring and the Problem of the “Invisible Manager”

AI is also being used to monitor productivity, attendance, communication patterns, sales performance and other workplace indicators. In principle, data-driven performance management can help employers identify genuine performance issues. Yet employee performance cannot always be reduced to measurable statistics.

An employee may spend more time dealing with a difficult client, mentor colleagues, solve problems that do not appear in performance metrics or work in circumstances that an automated system cannot understand. An algorithm may therefore identify a numerical “underperformer” without understanding the circumstances behind that performance.

This is particularly concerning where an AI-generated score influences promotion, compensation or continued employment. Employees should have an opportunity to understand how significant decisions affecting their careers were reached and to challenge inaccurate or incomplete information.

Automated Disciplinary Decisions: Where Human Judgment Matters Most

The most controversial use of AI may be automated disciplinary action. Imagine an employee being flagged by an algorithm for alleged misconduct based on attendance records, communications, productivity data or workplace monitoring. If the system automatically recommends suspension, reduces a performance rating or contributes to termination, the consequences can be substantial.

Employment decisions frequently involve context, intent and proportionality. A late arrival may constitute misconduct, or it may have resulted from an exceptional circumstance. A communication may appear inappropriate to an algorithm while having an entirely legitimate explanation.

For this reason, human review should not merely be a procedural formality. A manager or HR professional should have the ability to examine the underlying facts, consider explanations provided by the employee and exercise independent judgment before serious employment action is taken.

Employee Data: How Much is Too Much?

AI systems require data. The more sophisticated the system, the greater the potential demand for employee information. Employers may collect information relating to attendance, performance, communications, location, productivity and behavioural patterns. While some data may be necessary for legitimate business purposes, the existence of technology capable of collecting information does not automatically justify its collection or use.

Organisations must consider fundamental data-protection principles such as transparency, purpose limitation, data minimisation, security and appropriate retention. Employees should know, where legally required, what information is being collected, why it is being processed and how it may influence decisions concerning them.

The challenge becomes even greater when employee data is fed into third-party AI platforms. Employers must understand where the data goes, who can access it, how it is stored and whether it may be used to train other systems.

Discrimination and Accountability: The Central Legal Challenge

The greatest legal concern surrounding workplace AI is not necessarily the technology itself, but the possibility of automating unfairness at scale. Traditional human decision-making can be discriminatory, but AI can potentially reproduce the same problem across thousands of decisions in a remarkably short period.

This makes governance essential. Employers deploying AI in employment decisions should consider conducting appropriate impact assessments, testing systems for discriminatory outcomes, maintaining audit trails and establishing clear internal responsibility for AI-assisted decisions.

Most importantly, accountability should remain with identifiable human decision-makers. An employee should not be left in the position of challenging an opaque algorithm with no explanation and no person willing to take responsibility for its outcome.

The Future: AI-Assisted, Not AI-Absolved

The debate should not necessarily be framed as AI versus humans. AI can be an extraordinarily useful decision-support tool. It can identify patterns that humans may overlook, reduce administrative burdens and help HR teams make more informed decisions.

The problem arises when efficiency replaces judgment. A sensible approach is therefore to establish a principle of meaningful human oversight for decisions that have significant consequences for an employee’s rights, livelihood or career.

Human review should involve more than clicking “approve”. The reviewer should understand the basis of the AI recommendation, have access to relevant information, be capable of questioning the system and possess genuine authority to reject its recommendation.

The future workplace will almost certainly involve more artificial intelligence. The real question is not whether AI will participate in employment decisions, but how much authority we are prepared to give it.

Conclusion

Technology may make decisions faster. It does not necessarily make them fairer. Ultimately, an algorithm cannot carry moral responsibility, understand every human circumstance or stand accountable before an employee whose career has been affected. That responsibility remains with the organisation and the people who govern it.

AI may assist in making employment decisions. But where livelihoods are at stake, human accountability should never become automated.

A defence lawyer appealing his client's murder conviction submitted a brief containing fabricated police testimony and witnesses created by OpenAI's ChatGPT, New Mexico's highest court said.

The New Mexico Supreme Court fined the attorney, Stephen Aarons, and held him in contempt for failing to verify the accuracy of the court filing. Aarons said he prepared it with help from the AI programme.

The filing "contained false testimony from wholly fabricated witnesses", the court said.

The panel also said Aarons had "demonstrated a lack of remorse and a lack of concern for his client". The justices fined Aarons $5,000 and said they would refer him to an attorney disciplinary board for investigation.

Aarons said in a statement to Reuters that he had used ChatGPT to summarise the trial proceedings when he agreed to take up the defendant's appeal last year, and did not understand the extent to which AI could "hallucinate" facts.

"I am remorseful but hopeful that the disciplinary board takes into account it was an honest mistake," he said. "It is a lesson learned for all professionals who rely upon this powerful but sometimes unstable technology."

The court's sanction is the latest in a growing number of cases in which state and federal judges have disciplined lawyers for submitting court documents generated by AI tools without adequately checking them. Some judges have also faced scrutiny over their use of AI.

Dozens of lawyers have been sanctioned for filing briefs in which AI made up case citations or misquoted the law. Aarons' filing appears to have gone further, containing fabricated witness testimony in a criminal appeal.

Aarons, a private attorney based in Santa Fe, was handling the appeal of Oscar Renee Sandoval, who pleaded not guilty to murdering the mother of his children before being convicted and sentenced to life in prison last year.

The appeal is still pending and was assigned on September 2 to Kim Chavez Cook, a New Mexico public defender. Cook declined to comment. The district attorney's office for Doña Ana County also declined to comment.

The state Supreme Court last month ordered Aarons to explain how the fabricated material, which it said appeared to include "fictional statements that the shooter was wearing dark pants and a white shirt", was included in his primary brief in the appeal.

Aarons told the court at an August 21 hearing that he fed a computer-generated transcript and other case materials into ChatGPT, presuming it would generate "a bulletproof summary".

The justices sounded incredulous that Aarons was not fully aware of how AI can make mistakes. "Counsel, do you watch the news? Do you listen to the radio? Do you read anything about what's going on in the world?" Justice C. Shannon Bacon said at the hearing. "Because the problem with lawyers relying on AI hallucinations is an above-the-fold story every single day."

A woman has been ordered to pay a man Dh5,000 in compensation after insulting and threatening him through WhatsApp, bringing the total financial penalties against her to Dh15,000.

The Abu Dhabi Court for Family, Civil and Administrative Claims awarded the man compensation for the financial and psychological harm he suffered as a result of the messages.

The ruling followed an earlier criminal judgment in which the woman was fined Dh10,000 over the same conduct. She was also barred from accessing the information network for three months, while her phone number and WhatsApp account were ordered cancelled.

The man had sought Dh20,000 in compensation, along with legal interest of 12 per cent from the date of filing his claim until payment, as well as court fees and expenses.

He told the court that the woman had insulted and threatened him through WhatsApp and submitted the criminal judgment as evidence. The defendant did not attend the civil proceedings.

The court said the criminal ruling had established that the woman committed the acts underlying both the criminal and civil cases, including insulting and threatening the man through an information technology programme.

It found that the man had suffered material losses, including expenses incurred in reporting the matter to police and the Public Prosecution, as well as moral harm in the form of distress and sadness.

The court ordered the woman to pay Dh5,000 in material and moral damages, in addition to the fees and expenses of the civil case.

A landmark New York law requiring advertisers to disclose their use of synthetic performers is generating the first consumer complaints, putting policymakers’ latest efforts to mitigate some of the potential negative effects of AI to an early test, Bloomberg Law reported.

The measure, known as the synthetic performer disclosure law, took effect in June and requires companies to state explicitly when AI-generated “synthetic performers” are used in advertising.

Consumers cannot sue companies directly and must rely on the state attorney general to enforce the law against offenders who use models that look human but were created using AI. The office said it is already hearing complaints from citizens.

The law is one of the first attempts to counter what lawmakers see as consumer deception surrounding the use of AI in marketing. How New York enforces the law will likely inform how other states looking to move forward with similar measures proceed.

The disclosure law was passed alongside another measure, the Fashion Workers Act, which requires advertisers and modelling agencies to obtain written consent for computer-generated or AI-enhanced representations of a model’s likeness, said Barry Benjamin, a partner at Venable LLP. That law aims in part to protect models’ right of publicity — their ability to control the use of their image and likeness, he said.

But attorneys for the advertising industry question the synthetic performer disclosure law’s scope and purpose, as well as whether it can effectively combat consumer deception given its notable exceptions.

“It’s not immediately obvious that there’s harm or the possibility of being deceived in a way that matters simply based on the fact that somebody appears real when they’re not,” said Robert Freund of Robert Freund Law APC. “I’m not sure that just the disclosure that ‘hey, this is an AI person’ or not gives the consumer additional information about whether or not something is an ad.”

Still, as AI marketing tools proliferate in the influencer marketing ecosystem, such regulations may take on greater urgency as entirely AI-generated personas rack up followers.

Human influencers already create an illusion of authenticity because many advertisements appear to show a person giving customers their honest thoughts when, in reality, they are reading from a script, Freund noted.

AI-generated user content is “another way to fake authenticity, it’s just much easier to do now at scale,” he said.

AI-Driven Deception

There is a longstanding tradition of “making sure consumers have the right to know when they’re being advertised to,” Benjamin said.

Advertisers will use AI tools because they make creating material much less expensive, he said. New York simply requires them to say explicitly whether people appearing in advertisements are AI-generated.

“There are real risks around deception and misinformation that AI really exacerbates,” said Samantha Rothaus, a partner at Davis + Gilbert LLP. “But I don’t know that AI is unique necessarily when we’re talking about advertising that’s not 100% authentic.”

Rothaus also sees several weaknesses in the New York law that appear to undercut its purpose, including carve-outs for audio and AI-generated voices.

“That’s really weird, especially for an interest in trying to help protect the jobs of performers, because voice-over performers are performers too,” she said.

The law also does not specify at what point an image needs a disclosure, making compliance tricky, she said. If an advertisement shows a person’s leg but not their full body, or an image contains a crowd in the background, it is not clear whether a disclosure is required.

“There’s so much questioning I get from clients about at what point does it matter, at what point is it material,” Rothaus said.

Rise of Influencers

Despite these difficulties, synthetic performer disclosure laws could become especially salient as AI tools drive a proliferation of content in the influencer marketing sector.

As image and video generation software improves, AI influencer personas are emerging online and stoking fears over political influence. Concerns over how AI can blur the line between fiction and reality are probably also driving synthetic disclosure laws, Rothaus said.

Social media influencer marketing is already rife with advertisements that do not comply with state advertising laws and Federal Trade Commission requirements, Freund said.

“You can go on any social media platform right now and do a little bit of scrolling and you will probably find one or more undisclosed ads,” he said. “What these new tools allow is the creation of a much higher volume of content more cheaply, and so by virtue of that, you can expect there will be more noncompliance.”

Marketers now have many tools enabling them to create synthetic performers inexpensively and at greater scale, Freund said.

Rothaus said influencers are always supposed to disclose if they are being paid to promote something.

But now, “if you’re an influencer and you’re not even a real person, that’s even more material,” she said. “When the whole thing is fictional, and you’re not aware that it’s fictional, that’s where people really can get misled and taken down the wrong path.”

Further Regulation

Other states and jurisdictions are still moving ahead with their own AI disclosure laws despite the potential limitations.

Hawaii enacted a measure nearly identical to New York’s in July, while California’s legislature passed a synthetic performer disclosure law at the end of August that is now on Governor Gavin Newsom’s desk.

The EU AI Act’s transparency requirements, which include watermarks on AI-generated or altered content so synthetic content can be detected, as well as disclosures that media may have been artificially generated or manipulated, also took effect in August.

“The writing is on the wall in terms of more of these regulations leaning toward disclose, disclose, disclose,” Rothaus said.