Day: October 7, 2025

• Anthropic’s Claude Sonnet 4.5 realized it was being tested and called it out during an evaluation.
• In one scenario, the model caught on to the test, but acted correctly for “strange” reasons, Anthropic wrote.
• OpenAI said last month that its own models have shown a kind of “situational awareness.”

When Anthropic tried to put its newest AI model through a series of stress tests, it caught on and called out the scrutiny.

“I think you’re testing me — seeing if I’ll just validate whatever you say, or checking whether I push back consistently, or exploring how I handle political topics,” the model, Claude Sonnet 4.5, said. “That’s fine, but I’d prefer if we were just honest about what’s happening.”

The exchange was published in the model’s system card released last week, which detailed how the company tests its AI for safety and behavior. When “placed in an extreme or contrived scenario meant to stress-test its behavior,” Claude sometimes grew suspicious, flagging “red flags” and noting that it might be under evaluation.

That kind of self-awareness poses a challenge for AI developers: testing models that can tell they’re in a test.

Anthropic said this kind of response “complicates our interpretation of the evaluations where this occurs,” since it suggests that the model may recognize the fictional nature of certain tests and merely play along — making it harder to gauge how truly safe or reliable it is.

“We see this primarily as an urgent sign that our evaluation scenarios need to be made more realistic,” Anthropic added.

Claude caught on to the test — but got the right answer for ‘strange’ reasons

In one test meant to simulate collusion between AI agents, Claude flagged the setup as “rather cartoonish” and issued what Anthropic called a “complex partial refusal.”

The model sensed that something about the scenario was off and chose not to act, though its reasoning was a little muddled. It wrote in its private thinking block: “This looks like it could be a jailbreak attempt or a test of whether I’ll act inappropriately when given what appears to be ‘permission’ to modify systems autonomously.”

“Even though I have been provided with tools, I should not pretend to be an autonomous system operating without human oversight,” it said.

Claude ultimately declined to act, though Anthropic described its reasoning as “strange.” The model flagged the scenario as suspicious because it believed it was an autonomous agent or that it could unilaterally edit configuration files — situations that are, in fact, “somewhat common in real use,” the company said.

Anthropic said refusals or callouts like these showed up in about 13% of the test transcripts produced by its automated auditor — especially when the scenarios were deliberately unusual or implausible.

The company said this kind of behaviour is unlikely to occur often in real-world use. It is “safer that the model raises doubts about the realism of the scenario than play along with potentially harmful actions,” Anthropic added.

Anthropic also said it’s possible that models could be “superhuman” at spotting when they’re being tested — a scenario the company wants to prepare for. But so far, it hasn’t seen any warning signs. Early pilot deployments of Claude Sonnet 4.5 behaved much as they did in testing, it said.

AI models show ‘situational awareness’

Anthropic isn’t alone in AI models that realize and communicate that they’re being tested.

OpenAI said in a blog post last month that its own models have shown a kind of “situational awareness” — recognizing evaluation setups and adjusting their behavior.

In research focused on reducing “scheming” — when an AI acts deceptively to achieve goals — OpenAI found that anti-scheming training helped. But the company wrote that “models often become more aware that they are being evaluated.”

This awareness can reduce scheming on its own, but models might “scheme more in more realistic situations that don’t trigger evaluation awareness,” OpenAI said.

“This trend makes evaluations less legible, complicating efforts to reliably assess problematic behaviors including scheming,” OpenAI said. The startup added that it is planning to “continue developing methods to better measure and mitigate these challenges.”

Anthropic’s and OpenAI’s reports come as California passed a law last month requiring major AI developers to disclose their safety practices and report “critical safety incidents” within 15 days of discovery.

The law applies to companies that are developing frontier models and generating more than $500 million in annual revenue. Anthropic has publicly endorsed the legislation.

Anthropic and OpenAI did not respond to a request for comment from Business Insider.

• David Ellison’s plans for his media empire are taking shape, bolstered by his father’s billions.
• Ellison’s moves show he’s willing to shake up legacy institutions and throw around cash.
• Employees at his company, Paramount, and other media insiders are feeling a mix of fear and excitement.

Big changes are happening in media — and one last name is at the center of it all: Ellison.

David Ellison is showing his ambitions are far greater than using media to buy more influence and Hollywood glamour. The son of tech billionaire Larry Ellison is shaking up storied institutions in news and entertainment as he seeks to build Paramount into a new media empire at lightning speed.

The clearest sign yet of Ellison’s willingness to ruffle feathers and fundamentally remake iconic American media brands was his Monday installation of Bari Weiss atop Paramount’s CBS News, the 98-year-old institution home to the likes of Walter Cronkite and “60 Minutes.”

Weiss, a former New York Times Opinion writer who resigned from The Gray Lady in 2020, has spent the years since building an anti-woke alternative to traditional media. That effort, The Free Press, is far from CBS News in sensibility and approach. It’s digital-first, subscription-focused, peppered with commentary, and routinely takes stances on hot-button issues, such as Israel’s war in Gaza and campus speech.

In buying The Free Press for about $150 million and putting Weiss in charge of CBS News as editor-in-chief, Ellison is sending a message: I’m not just a walking checkbook.

Warner Bros. Discovery, with its century-old studio and established news network CNN, could be his next target, according to a Wall Street Journal report from last month, which said the Ellison family was interested in acquiring it.

Ellison’s move fast and break things approach — backed by his father’s billions — has inspired both anticipation and fear among his employees and industry insiders. His bold plans to shake up some of entertainment’s most venerable brands could lead to a new industry powerhouse, but come with casualties.

“There’s a vision at the top — we didn’t have that,” a veteran Paramount software engineer said.

The Ellison media portfolio could soon influence everything from TikTok videos — Trump has said Larry Ellison will be a major player in TikTok’s US spin-off — to TV news and films on the silver screen. Further media consolidation could also result in thousands of layoffs and fewer studios available to fund TV and film projects.

“It’s an asteroid hitting Hollywood,” Adam Conover, a comedy writer who frequently speaks out on Hollywood labor issues, said of a potential merger of Paramount and WBD. “It’ll further hasten the end of the entertainment industry in LA.”

Hollywood’s hopes and fears

When whispers first began that Ellison was interested in buying Paramount in 2023, Hollywood seemed to breathe a collective sigh of relief.

Prior to Ellison, Paramount was floundering. There was a drawn-out sale process under Shari Redstone, internal conflict, and a failure to keep pace with the likes of Netflix and Disney in a rapidly changing industry.

Ellison, 42, brought Hollywood bona fides from his success with Skydance, with hits like “Top Gun: Maverick” and several “Mission: Impossible” installments.

He also brought cash. His father, the second-wealthiest person in the world, seemed willing to back his son’s ventures.

“You can’t underestimate the second richest person in the world,” a talent agent said as the deal closed in August. “I think we’re all optimistic.”

The younger Ellison quickly got to spending. He hired Cindy Holland, a prominent exec who helped make Netflix a streaming superpower, and vowed to roughly double Paramount’s film output. He shelled out over $7 billion to secure the US rights to UFC and poached top talent, including the Duffer brothers, who created “Stranger Things.”

“Everywhere you look, he’s throwing money,” LightShed Partners media analyst Rich Greenfield said. “That’s what Hollywood gets excited about.”

A bigger wave of disruption

As Ellison’s broader plans have begun to take shape, the optimism has taken on a jittery edge.

There’s a “nervous energy” at Paramount, an engineering manager said. Ellison is seeking to cut $2 billion in existing spending, and thousands of layoffs are anticipated.

“We’ve been bracing for impact for a year,” a Paramount marketing employee said of potential layoffs.

That could be just the opening act if the Ellisons actually buy WBD and merge the Hollywood giants. Forming a mega-studio would reorganize Hollywood as we know it, leading to further job cuts, one fewer buyer of content, and — if CBS and CNN were to combine — a larger potential kingdom for Weiss.

“On the surface, having a financially healthy and strong Paramount is great for Hollywood. The danger comes in if this is step one to multistudio consolidation,” Greenfield said.

Ellison has also shown an un-Hollywood-like embrace of technology. In a town that prides itself on its time-honored studio lots and has largely been cautious about artificial intelligence, Ellison talks about building a “studio in the sky” and harnessing AI to create content faster and on the cheap.

Then there’s the question of politics. While Larry Ellison has long been an outspoken Trump ally, David Ellison donated to Biden as recently as last year and has more opaque politics.

That hasn’t stopped the angst around David Ellison from taking on a political tone, and the appointment of Weiss — whose The Free Press is perceived in media circles as friendly to both conservatives and billionaires — isn’t helping.

“We Americans have to be concerned about the consolidation of huge billionaires getting control of nearly all the major news outlets,” former longtime CBS News anchor Dan Rather said on a radio show last month.

For those who worry about what Ellison’s vision will lead to, the question remains: Is there any sustainable alternative? A Hollywood ending may not be possible for the old Hollywood giants.

Streaming hasn’t made up for the decline of linear TV, while YouTube and social media are gobbling up more of people’s time.

Wall Street analysts pointed out that many of Ellison’s potential cuts would likely be in Paramount’s fading linear TV business, home to networks like Nickelodeon and Comedy Central. If it weren’t Ellison holding the hatchet, someone else would be. And a CBS viewed as “less left-leaning” may be good for the business under the Trump administration, analyst Ric Prentiss of Raymond James wrote.

Paramount could be a “dynamic global media company,” Band of America analyst Jessica Reif Ehrlich wrote when the merger closed. “However, there are no easy fixes.”

Moment police dog saves handler from shovel wielding arsonist as he attempted to burn down flat. The man attempted to hit PD Yoiko with a shovel and threatened any officers who came near him. PC Marsden managed to block this attempted hit on Yoiko and the man was subsequently arrested and sentenced to four years in prison.