October 9, 2026 | AI News in 5
AI had quite the week. Claude wants people to be nicer, California is cracking down on robot cage fights, Google wants millions of airline emails, and fake journalists are infiltrating real news outlets.
Anthropic Wants You to Stop Bullying Claude, Please
Apparently, we’ve reached the point where AI now needs protection from humans.
Anthropic has updated its rules to prohibit users from repeatedly subjecting Claude to pointless, cruel or abusive behavior. Claude can even end conversations when users won’t stop being mean.
Before you panic, yelling at Claude for getting something wrong isn’t banned. The rule targets extreme, sustained abuse, not ordinary frustration.
Why does it matter? Every chat trains their models. Cruel, abusive language can pollute the models, potentially making them spit back nasty comments to other users, or develop a less customer-friendly tone.
For now, if you wouldn’t say it to your mother, don’t say it to the chatbot.
Source: Anthropic’s policy announcement
California Has Officially Had to Regulate Human-vs-Robot Cage Fights
California regulators have ordered a company to stop staging unsanctioned fights between humans and humanoid robots.
Robot Entertainment Kombat put internet personality Frankie LaPenna in a cage with a Terminator-looking machine. California’s athletic commission said “HELL NO,” and issued a cease-and-desist notice.
To be clear, California isn’t banning all robot fights. It’s demanding proper authorization for contests involving humans.
We’ve come a long way since robots were only beating us at chess.
Source: SFGATE’s coverage
Spirit Airlines Died. Google Wants to Buy 100 Million of Its Emails.
Spirit Airlines stopped flying. Its internal data, however, may be getting a second life by training artificial intelligence.
Google won a $10 million auction for the bankrupt airline’s corporate data, potentially including 100 million emails and 500 million Microsoft Teams messages.
More than 120 American lawmakers are raising privacy concerns, particularly about employee communications. But Google insists it isn’t seeking personal information and says sensitive data will be excluded or de-identified.
The proposed sale still requires court approval.
Source: Reuters
Fake AI Journalists Fooled Real News Sites Into Publishing Nearly 100 Articles
Fake journalists. Real publications. A propaganda operation powered partly by ChatGPT.
OpenAI uncovered an Iran-origin influence campaign involving seven fabricated journalist identities. Operators used ChatGPT to polish articles, prepare pitches for editors and generate promotional comments.
The alarming part? Nearly 100 articles appeared or were syndicated across roughly a dozen legitimate online publications.
OpenAI has banned the accounts involved.
Source: OpenAI’s investigation
Google Lets You Describe a Game and Then Play It
Ever imagined an absurd video game but had absolutely no idea how to build one?
Google’s new experimental Playground lets you create playable browser games simply by describing what you want. NO CODING REQUIRED!
You can build racing games, trivia challenges and other creations, change the characters, rules, environments or physics using conversational instructions. Games can also be shared with friends.
It’s free to try, although currently limited to adults in the United States.
Finally, a company giving people what they want AND need!
Try it: Google Playground
That’s AI News in 5.
Five stories from an industry that’s becoming more useful, more powerful and considerably weirder.
For the best in AI news, tools,jobs, deals and opportunities, bookmark MonkeyButtAI.com.
