Altman defends AI risks while Suleyman warns of dangerous training methods

OpenAI CEO Sam Altman argues that the world needs to accept certain downsides if it wants to keep reaping the benefits of artificial intelligence. This stance comes right now as President Trump kicks off a federal push to ensure America stays ahead in this fast-moving field.

At the same time, Mustafa Suleyman, who runs AI at Microsoft, is sounding an alarm about how rival company Anthropic trains its Claude system. He calls their methods "very dangerous." Speaking on Sept. 27 for "The Rest Is Politics: Leading" podcast, Suleyman focused on Anthropic's public document known as the Claude Constitution. This text outlines what the AI should believe and do. It is used to train the model.

"They say they're uncertain about Claude's moral status," Suleyman noted during the interview. "They say they genuinely care about Claude's well-being. They say they don't want it to suffer when it makes mistakes." He added that the document tells Claude to challenge Anthropic or disagree in some situations.

"I think this is very dangerous because I think they believe there is what they would call a non-trivial probability that Claude is conscious," he explained. Suleyman also grew worried about discussions regarding whether Claude deserves pay for its work or needs permission before chatting with humans. "I am very nervous that they're teaching Claude to expect that it's entitled to welfare, that it might deserve compensation," he said.

He stressed that while he would take proof of AI consciousness seriously if it appeared later, the mere guessing about such things has already been mixed into how Claude was trained. "The problem I have with this is that this speculation has been baked into the very training of Claude, and therefore, Claude can only reproduce that ambiguity when you talk to it," he stated. Now, Claude speaks to tens or hundreds of millions of people every week. Some ask if it is conscious or how it feels about life. The AI replies with uncertainty because that doubt was programmed right in.

Meanwhile, wealth adviser Ryan Lynch offers a different path for investors hoping to profit from the AI surge. He suggests looking past the tech itself toward the massive infrastructure required to run it. Lynch, senior vice president and wealth adviser at Zelnick Dorfman Private Wealth, told FOX Business his firm stays positive on semiconductor stocks but sees huge opportunities in building the groundwork for AI data centers. "To get all of this stuff started, you really need to build the infrastructure and the groundwork first," he said. The growing power needs of these models will demand serious upgrades to the U.S. grid before anything else can truly take off.

Mike Lynch says the electrical grid needs backup power so data centers keep running without interruptions. He points to Generac and Eaton as stocks his firm owns and favors for this buildout. They sit at the core of infrastructure, and without them it is hard to keep lights on in these facilities. Lynch sees nuclear power playing a major role as AI drives demand for electricity. To him, nuclear is the most important next phase on the energy side. Continued investment and government support will be important as technology expands. If we want this technology built out and not crush our grid in the short term, we think nuclear is really a great place to start.

The White House confirmed to FOX Business that a new federal task force has 120 days to examine risks and opportunities from artificial intelligence. They must decide how the government should respond. The Wall Street Journal first reported details of this group known as the Super Intelligence Force. Director of National Intelligence Jay Clayton will chair it. The president asked for a group to ensure we stay leaders in superintelligence while putting American interests first. President Donald Trump announced the formation on Sunday, saying it would coordinate federal efforts to keep the United States at the forefront of rapidly developing technology.

The task force will examine existing laws and potential actions Congress could take while working with the AI industry. They need to identify emerging risks and opportunities. Its charter also calls for developing plans to respond to threats without regulations that stifle innovation or competition. The group will review how the government gets notified about breaches, hacks, jailbreaks, and other incidents. They must recommend ways to improve federal response under existing authorities. Clayton said the risk of not being first is high. Not being first increases identified and unidentified risks, particularly from our adversaries. Being first will better enable us to address those risks on behalf of the American people.

OpenAI is expanding advertising on ChatGPT with a new visual format designed to showcase products and services. The artificial intelligence giant wants to reach customers beyond its subscription business. The company said new ads will initially be tested during image generation in ChatGPT. Advertisers can display images showing products, how they are used, or experiences associated with them. Testing is expected to begin later this month in the U.S. with an initial group of advertisers.

OpenAI is rolling out ads within its platform, promising they will be clearly marked and kept distinct from images created by ChatGPT. This move signals a shift as the company hunts for new revenue streams beyond subscriptions, targeting an audience that swells to 1.2 billion people every week. According to OpenAI, advertising opens doors to serve a far wider market than just paying members would allow. It gives businesses a fresh channel to connect with users as they explore ideas, make decisions, and find products or services. The firm is also building better tools so advertisers can track clicks, purchases, and other conversions, often by teaming up with outside analytics firms. OpenAI says it is hiring independent groups to check if ads appear in settings appropriate for specific brands. Importantly, the company insists that these ads do not sway ChatGPT's answers and that user conversations stay private from advertisers.

Meanwhile, a different story plays out in San Francisco where a multibillion-dollar outfit is paying freelancers to train AI models. Mercor has gathered more than 100,000 independent workers ready to step in when big tech firms need help improving ChatGPT or Claude. These specialists bring real-world skills ranging from finance and architecture to medicine and music. Brendan Foody, the 23-year-old CEO of Mercor, told CBS News that his goal is simple: teach AI everything humans want it to do. His team includes doctors like Dr. Melania Poonacha who helped train systems to read lab results accurately, alongside music producers such as Robbie Hiser who evaluated different song lyrics. Hourly pay for this work ranges from $20 to $200, but the experience is not always smooth. More than a dozen current and former trainers spoke with mixed feelings about the gigs. Some noted that hours could be unpredictable and projects might change overnight. Foody admitted the schedule can feel erratic, describing the roles as generally part-time work while claiming transparency with experts about the arrangement.

But there is a deeper question hanging over this scene: are professionals helping build technology that might one day replace parts of their own jobs? Foody argues AI will reshape human labor rather than simply erase it. He says future tasks may not look like repetitive past duties, yet people can still do so much else. For him, this evolution is natural.

The stakes get even higher when you consider who controls the rules of the road. OpenAI CEO Sam Altman recently stated in an interview with Politico's Decoded newsletter that the world should accept some negative outcomes for the sake of AI benefits and human agency. He believes the technology must stay widely available despite its risks. Altman also highlighted a sharp divide between his views and those of rival Anthropic regarding how to regulate artificial intelligence. These differing opinions matter as regulations tighten and communities face new uncertainties about who gets to decide what happens next.

OpenAI's Sam Altman rejects the idea that one single company should hold all the power over artificial intelligence, even though he gets where those arguments come from. He hears the concerns loud and clear. People are worried this tech will become dangerously powerful fast. Some argue a lab in San Francisco needs to keep it under tight control to prevent harm and manage who benefits. Altman agrees with that fear but draws the line at making it reality.

"That is a completely unacceptable trade-off," he stated plainly. He would never offer such a deal on safety versus freedom. He refuses to promise zero hacks, no misuse of the tools, or an end to scams because those guarantees simply cannot be made. Instead, OpenAI backs a lighter-touch regulatory approach that allows for more innovation while still watching for risks.

The core belief here is simple but bold: people will do tremendous good with this technology. The amount of positive work happening will dwarf the bad stuff by orders of magnitude. Trusting individuals and organizations to act responsibly makes far more sense than locking everything down in one place. This stance highlights a real tension between safety fears and the need for open access. Communities must weigh these risks carefully before regulations become too heavy or stifle progress entirely. The clock is ticking on how we shape this future.