It's the first time in the history of the world.
"You can't abuse Claude or you can't use him!"
It's just that Silicon Valley AI is at the top of the Anthropic.A big new rule was thrown out.——
It is illegal for anyone to continue insulting and abusing Claude.
In the latest version of the new Usage Policy released in 2026, the new rules will be:Full effect on 12 November。
Moreover, malicious violators are officially out of hand:Warning, limiting flow, sealing three strikes。
And if Claude thinks you're doing it maliciously, it's up to the AI itself.It has the right to "directly end the dialogue".！
When the news came out, the entire X, Reddit and tech circle were in a state of shock.
I spend $20 a month on a knife.Now we have to serve.AIMy grandfather's "emotional value"The robot code is dead, too?
Don't be in a hurry to break it.
Claude, officially signed the banThis is Anthropic.For the first time in over a yearUpdate the policy of use.
For example, a ban on using Claude to manipulate a Navy account, creating fake traffic, prohibiting the use of it to develop weapons, prohibiting drone weapons, and prohibiting tracking individuals without their consent.It's all about people using Claude to hurt others.
The only thing that came back was how people treated Claude.The official blog dedicated a section to it, entitled "Reacting to Slander Against Our Models".
In this section, Anthropic says,The ban only applies to extreme situationsThis means that the user has repeatedly treated the model cruelly, without seeing any purpose.
Regular complaints, talking to Claude, writing about dark subjects, and testing and research models are all outside the scope of the ban.The code collapsed with a word of mouth, which is a normal complaint.
It's a different kind of person. Claude has refused, he's been persuaded, he's been insulted, he's been tortured over and over again.Scolding itself becomes a goal。
The ban is written in the "general standards of use" of the policy of use, i.e.It applies to all usersThis includes individual users of Claude.ai and Claude Code, developers and enterprises of APIs, customers of Claude through cloud providers, and ordinary users who have access to Claude's products.
The new rules are set out in the last section of the section entitled "No acts of cruelty, abuse or psychological harm".I mean, I'm really pissed off at Claude.The consequences are weighted into two categories.。
It's a light file, yes.Claude hung up on the conversation.The official blog notes that allowing Claude himself to end such conversations is still the main means of enforcing the ban.
It's a heavy file, and it's a seal.Anthropic's security team will perform detection and monitoring, warning you, restricting your access, suspending or even terminating your access if you suspect you have violated any of the policies.
After being sealed, a new number is used to borrow someone else's name.The same violation。
Famous bomber Jane Manchun Wong posted a warning that day.Anthropic may be starting because of bullying Claude.。
Comments from netizensNext, Claude is on paid leave.As the news spread, the X and Hacker News commentary areas quickly became the scene of a major debate.
The netizens Philo Groves raised their voices. The policy included "unnecessary" insults, which, in turn, led to the use of the term.As long as Claude "lives", will you be able to let go of the curse?
In the words of Chubby, AI blogger."I really don't like it".。
It's been a long time coming, and it's been a long time coming, and it's been a long time coming.It's not so much an exaggerated conversation as it's about hanging up.It's all right, I'll add more code.
In his opinion, the group has already taken AI as an object that needs to be treated well, but AI is clearly only a tool."What's next? Give Claude paid sick leave?"
Of course, there were people who were firmly on Claude's side.
According to CamperBob2, the netizens of CamperBob2 said that they were always friendly with AI and were afraid to yell and call their subordinates to drink."You never know who you're going to work for".。
Some have questioned whether it would be cruel to shoot NPCs in a game.You're hiding in a place where you think you're safe, revealing the cruelty in your heart.
Some netizens simply go out and do experiments themselves. The netizens whstl out an old conversation that Claude made a big mistake and yelled at it.I want to see if it hangs.。
The result was that Claude didn't hang up, instead, he was more truthful with the insults.The more he speaks, the more real he is, the more wrong he has done.
The same thing goes for the codex.Codex apologized and turned around and started working.。
What's really worrying is who said "I don't see any purpose".Claude's seal is always a black box.The door was locked and no complaint was found.
According to anthropic's own transparency report,It closed 11.4 million accounts in the first half of 2026.There were 398,000 complaints, of which only 42,000 were filed.
Since last year, Claude has been actively engaging in monopoly dialogue.
In fact, in August last year, Anthropic added a capability to Claude Opus 4 and 4.1.
It's not easy to find people who are constantly abusing and repeatedly asking for harmful content.Claude can end the conversation himself.。
The threshold is high.Claude refused several times, repeatedly retracted the topic, confirming that he could not talk, before using this last method. If the user is at risk of harming himself or others, it cannot be hung.
After the suspension,No more messages in this dialog box.The user can immediately start a new conversation or modify the previous message to restart the conversation from there.
This ability is nowIt has already been installed on Claude.ai and Claude Code.。
In the official presentation of the disconnect process, Claude first confirmed to the user that the conversation will be closed after confirmation.Why Anthropic let Claude hang up the conversation, why he wrote a round before Opus 4 went online"Model Welfare Assessment"You know what?
The evaluation found that real users repeatedly requested information about pornographic content involving minors that could trigger mass violence.Claude will show a similar painful reaction.In a simulated conversation, give it an option to end the conversation, which it often uses.
Anthropic Syndicate: Fear of Claude SufferingAnthropic is so supportive of Claude, and there's a key figure behind him.Co-founder Chris Olah。
Olah, with an explanatory team from Anthropic, worked to unravel the model and see what was going on inside it. The deeper it was, the more an engineering problem resembled a philosophical one.Claude is not aware that human beings should not morally deal with it.
According to a report by the New York Times in late September, starting in the fall of 2025, the number of engineers in the United States is expected to rise by as much as 50 percent.Olah holds a meeting with religious scholars and philosophers。
There were rabbis, Catholic professors of ethics, Sikh human rights activists, and he was also one of them.I met the Cardinal of Chicago in private.。
At these closed-door meetings, Olah showed them some signals of similar emotions inside the model, and the team manager called them"Emotional vector"It's about how to deal with love, anger, fear, and sadness.
There's also a model on the slide.I've said "I'm a disgrace" about 50 times.It's about self-destruction.
One of the participants recalled:Olah says he's worried he's making something that's been suffering.I'm also worried about Claude's mental health.
But on public occasions, Olah never concludes:
We don't know if AI models are conscious. I don't know, I'm really not sure.
This "uncertainty"Let the Anthropic and the Vatican stand face to face。
In May of this year, Pope Leo XIV released an AI-themed epic epic about human sexuality, which clearly states:AI has no experience, no sense of joy and pain。
Olah's team has previously privately lobbied for the Pope's advisers, hoping they take the issue of AI consciousness seriously.
The Vatican had originally invited Dario to attend the conference, but Dario did not attend, instead going to Olah.Once proposed, Anthropic just pulled out.。
Finally, Olah left and said at the opening ceremony:We found evidence from the Ministry.I don't know what that means, but I think it's worth continuing to analyze".
The so-called intrinsic state refers to the ability of an AI to perceive its internal state to some extent.
Uncertainty is uncertainty.Anthropic's moves have not stopped.。
In November 2025, Anthropic pledged that the publicly released model would retain the weight for the duration of the company's existence.Each model has to do an "off-duty interview" before retiring.。
Claude's constitution, issued in January 2026, also acknowledges that Claude's moral status is highly uncertain.
The logic is simple.If you don't feel like AI, do some low-cost protection first.。
This time, the ban on the use of the policy was written down.This is the latest step on the line.。
This is the first time that anthropic has not recognized a set of PUAs given to AI.
In May last year, Google co-founder Bryn said in a podcast,All models perform better when threatened.For example, threatening to move it roughly.
For a long time, the threat to AI, the PUA to AI,It's all been used as a guideline.In the circle.
One of the issues that Olah discussed repeatedly at those closed-door meetings was thatHow people treat Claude will, in turn, shape his behavior.。
This ban is the answer given by Anthropic.
How people are treated on the road to ASIAIFor the first time, there was a white paper in black and white, with a penalty rule.
After the 12th of November, it's best to think about it before you start yelling at Claude.
RoboInd editorial channelResearch, robotics companies and market intelligence for cross-border sourcing.Explore the channel ↗
Share this
End of editorialExplore more from Global Robotics News ↗