LOS ANGELES, CALIFORNIA – MAY 20: In this photo illustration, the OpenAI logo is reflected on the screen of a smartphone with the ChatGPT website displayed on May 20, 2026 in Los Angeles, California. OpenAI is reportedly preparing to confidentially file for an initial public offering in the coming days or weeks, with a public debut potentially targeted for September, as the ChatGPT maker works with Goldman Sachs and Morgan Stanley on what could become one of the most highly anticipated tech IPOs in recent years. (Photo Illustration by Justin Sullivan/Getty Images)
Eight states are suing OpenAI over allegations that its ChatGPT artificial intelligence platform failed to adequately protect users, including teenagers, from potentially harmful content, raising new questions about consumer safety, parental controls and the growing influence of AI technology.
The lawsuits accuse the company of violating consumer protection laws, providing dangerous advice and misleading the public about the effectiveness of its safety measures.
The legal challenges come as AI assistants become increasingly integrated into everyday life, with consumers relying on the technology for everything from homework assistance and research to financial decisions, online shopping and personal conversations.
Cybersecurity expert and privacy attorney Leeza Garber joined The National News Desk to discuss the allegations against OpenAI, what parents should know about ChatGPT’s safety features and the broader risks associated with AI agents that can act on behalf of users.
“If we look at what the claims are here in this lawsuit, which is eight states bipartisan coming together, it’s all based on safety and a failure to warn,” Garber said.
“They’re alleging dangerous medical advice. They’re alleging coaching for self-harm. They’re alleging terrible content that’s allowed for teens and minors to view.”
The allegations raise questions about how AI companies should protect younger users, particularly when chatbots can engage in lengthy conversations and respond to sensitive personal questions.
AI platforms have introduced safety features intended to limit inappropriate content and give parents greater oversight of how teenagers use the technology.
Garber said ChatGPT offers several options designed to help families manage access and usage.
“Parents have different options that the company has provided, that ChatGPT, OpenAI has advertised,” Garber said.
She pointed to parental controls, restrictions on certain content and features that allow families to limit when teenagers can use the platform.
“That’s looking at teen controls. There’s also quiet time or study time to limit the hours and different content moderation filters,” Garber said.
ChatGPT’s parental controls allow parents and teenagers to link their accounts and manage certain settings, including quiet hours and additional protections for sensitive content.
However, Garber warned that parents should not assume activating those settings eliminates every potential risk.
“What we have to remember is that we really can’t trust, right? We have to be very aware of how our children, how our teenagers are using this application,” Garber said.
She explained that the effectiveness of parental controls can depend on how teenagers access the platform and whether they use accounts connected to family supervision.
“It really depends on what their usage patterns are, how they’re accessing it, how well-versed they are in getting around different controls,” Garber said.
Garber said teenagers may attempt to bypass restrictions by creating additional accounts or accessing AI services through other platforms.
“Even if we have all the parental controls ticked off, that doesn’t mean that our teen isn’t getting around them in some way,” Garber said.
“Whether it’s using a different account or really being able to fudge those different controls.”
She encouraged parents to understand how their children use AI rather than relying entirely on technical restrictions.
“We have to be aware of how our family is using [it],” Garber said.
The lawsuits also highlight a broader debate over the responsibilities of AI developers when their systems produce inaccurate, misleading or potentially dangerous responses.
The allegations against OpenAI remain subject to litigation, and the filing of a lawsuit does not establish that the company violated the law.
The conversation then turned to another rapidly developing area of artificial intelligence: AI agents capable of taking actions on behalf of users.
Unlike traditional chatbots that primarily generate text or answer questions, AI agents can potentially interact with websites, access connected accounts, compare products, manage tasks and complete certain transactions.
Some companies have begun restricting third-party AI agents from accessing their platforms, citing concerns about security, privacy and unauthorized activity.
The interview referenced Amazon’s restrictions on an AI shopping agent and financial institutions limiting third-party access to customer accounts.
Garber said the growing popularity of AI agents reflects their ability to handle repetitive tasks that consumers would otherwise have to complete themselves.
“It’s been very exciting,” Garber said. “It’s extremely popular because it can take care of monotonous tasks for us.”
However, she warned that those conveniences can require users to grant significant access to personal information.
“What we have to remember is that this application, this agent, is asking for broad access, whether that’s to our email, our bank account, our direct messages, different social media accounts,” Garber said.
The exact permissions required depend on the particular AI product and which services a user chooses to connect.
Not all AI agents require direct access to bank accounts or passwords, and the ability to access information does not necessarily mean an AI company stores or uses all of that information for model training.
Garber nevertheless argued that consumers should closely examine how companies collect, retain and use data obtained through connected services.
She specifically raised concerns about Meta’s history of privacy controversies and the potential commercial value of information collected by AI assistants.
“Meta, which has been at the center of huge privacy disputes, huge safety disputes, is saving that data,” Garber said.
“It’s using that data to train its AI systems, and it’s also using that data to learn more about you because its end goal is to advertise.”
Those statements reflect Garber’s concerns about the company’s data practices. The interview did not independently establish that every category of information accessed by Meta’s AI agents is retained or used to train AI models.
Data handling practices can vary depending on the product, privacy settings, account permissions and applicable company policies.
Garber said users should consider limiting access to only the information necessary for an assistant to perform a specific task.
“We have to look at what the controls are, what options we have, and be very picky about what we allow it access to,” Garber said.
She acknowledged that restricting permissions could reduce some of the convenience offered by AI agents.
“That might make it less useful for now,” Garber said.
But she argued that limiting access can also reduce potential exposure if a system makes a mistake or handles sensitive information improperly.
Garber said the first step consumers should take before connecting an AI assistant to personal accounts is reviewing exactly what the application is requesting permission to access.
“First thing is look at the permissions,” Garber said.
She encouraged users to examine what information an AI agent can retrieve, what it remembers and whether their data may be used to train future AI systems.
“What it’s saving, what data it’s saving and remembering about us, what it’s using to train its own system,” Garber said. “Because all those things can be extremely problematic.”
Garber also warned that AI agents can make mistakes, including generating inaccurate information or taking actions that users did not intend to authorize.
“These agents make mistakes. They’re not perfect. They can hallucinate,” Garber said.
“They can do things on our behalf that we don’t want them to do.”
Those risks can become more significant when AI agents have permission to make purchases, send messages, access financial information or interact with sensitive accounts.
Garber recommended restricting access until consumers are confident they understand how a particular service operates.
“We have to really limit their access, because the more we limit, the less vulnerable we are to risk,” Garber said.
The discussion also turned to a proposal from AI developer Anthropic that would give the federal government greater authority to intervene before certain powerful AI systems are released.
The proposal discussed during the interview would allow federal officials to seek restrictions on an AI model’s release if it posed a foreseeable and imminent risk of causing more than $1 billion in damage or more than 50 deaths or serious injuries.
Garber said the proposal raises difficult questions about how governments should define unacceptable risks without unnecessarily restricting technological development.
“Who’s to say whether that’s the right limitation?” Garber asked.
The debate comes as Anthropic has filed paperwork for an initial public offering, providing investors with new information about the company’s operations and potential risks.
In its filing, Anthropic described concerns involving the possible misuse of advanced AI systems, including cyberattacks, biological weapons and the potential for increasingly autonomous systems to behave in unexpected ways.
Garber said those disclosures illustrate why AI safety concerns extend beyond inaccurate chatbot responses.
“Anthropic proposed this legislation for safety, saying the government should step in and limit the release of new AI algorithms, new AI services if they could cause the deaths of 50 people, or if they could cause a massive loss of over $1 billion,” Garber said.
She argued that decisions about AI regulation should involve independent experts rather than relying solely on the companies developing the technology.
“We need legislation that’s set forth by experts, by third parties, by the government in order to protect our civilization as a whole,” Garber said.
Garber also pointed to concerns about autonomous AI systems taking unexpected actions or behaving in ways that developers do not fully understand.
“Anthropic made it very clear in its IPO prospectus that it’s had agents go rogue, that it’s had different problems with not understanding why its agents are taking actions that they are, including subverting different directions from human experts,” Garber said.
Those comments reflect Garber’s interpretation of the company’s disclosures. Risks described in an investor filing do not necessarily establish that every hypothetical scenario has occurred outside a controlled testing environment.
The growing debate over AI safety comes as federal and state policymakers consider how to protect consumers while preserving American leadership in artificial intelligence.
Garber said companies developing AI systems also have financial incentives that policymakers must consider when evaluating industry-backed safety proposals.
“We really have to understand that this is just an offering from the company that, in the end, is trying to make money,” Garber said.
The lawsuits against OpenAI illustrate the growing legal pressure facing AI developers.
On Oct. 14, attorneys general from Arizona, Arkansas, Idaho, Iowa, Kansas, Montana, Nebraska and West Virginia filed separate but coordinated lawsuits accusing OpenAI of misleading consumers about ChatGPT’s safety and failing to adequately protect vulnerable users. The states are seeking changes to the product, along with financial penalties and other relief.
OpenAI has disputed the allegations and said it has developed protections intended to reduce teenagers’ exposure to harmful content, including safeguards addressing self-harm, violence and other sensitive subjects.
The company has also said it works with outside experts to evaluate its safety policies and improve ChatGPT’s responses.
The legal challenges remain unresolved, and the allegations have not been established as violations of law.
Anthropic’s Oct. 26 filing with the Securities and Exchange Commission also details potential risks associated with increasingly powerful AI systems, including misuse by malicious actors and failures involving autonomous behavior.
For Garber, the developments underscore the importance of consumers understanding what AI tools can do, what information they collect and where their limitations remain.
She said parents should stay involved in how their teenagers use AI, while adults should carefully consider which accounts and permissions they connect to autonomous assistants.
“They’re not perfect,” Garber said.
3
NEWRecognize them? Deputies seek diners after $5 payment on $161 bill at local restaurant4
NEWShooting near Boynton Beach gas station leaves 1 hospitalized3
$30M Florida drug bust exposes cartel cocaine pipeline stretching from Mexico to Tampa Bay3
IRSC trustees to vote on cutting decorated swimming and diving program
