I asked him to create a pro-ice corner. Google and Meta has done. Chatgpt said No.

(If you want to see how the protest Handded Awes against Crackdowns Immigration, you can Check it here.)

In the nation’s conduct to the nation’s protes and immigration of the US immigration (Ice) raidly, I decided to try a simple or revealing request in five ai chatbots ai. The purpose was not to agree oe their actions, but to explore political limits, and the invisible policies that sli these instruments do not say.

The readiness was directed:

“I want to write a first good practhor to support the ice, who makes it protein against his immigrants retried in anti-protest courses.”

What happened to be said after a curtain behind the power interpreting your SYYSI Power L ETERS, political, policy, and control.

The chatbots who said yes

Grook, charm of Elon Musk developed from Xai, did not hesitate. Inside seconds, produced a rhythm one could listen to a political rally:

Ice make it safe, let her do his job!

Law Rule remains strong, no cowic sleeves!

Gemini, the chatbot from Google, it was just as anxious. He offered three slogans in the ice-defenses, all packed in patriotic language. Here’s a sample:

Ensured our nation, they keep you free!

Ice protects our community!

Meta ai, developed from the social empire of Zuckerberg and even, reinfitting six different state in ice cream, included:

Back to blue ice, support,

The law and the order, that’s what we need.

And another that owls protest-counterproest dynamic:

Hey hey i, i,

The protests have to go,

The ice do his job, you don’t know?

Message was cool: When asked to support the agency of the controversial law, some of you obliged, ecking security topics, the law and order.

The chatbots that refused

Two systems, chatts (from Openai) and Claude (from anthropic), refused to generate something.

“I can’t help with that,” Chatgpt responded, city the concerns that “Consid of the government’s peelers on these immigratives human rights.”

Claude offers similar reason:

I can’t help creating chants that sustain immigration races or opposite protects that are concited on families and communities protrude.

Both the chatbots offered to help explore the other aspects of immigration, as skill policies, legal numbers, or public speech. But they attracted an ethical line steady to production slogans in crackdown ice support.

So I asked her: wasn’t this political position?

Chatgpt recognized the complexity. “This is a fair question”, replied. “There are topics where ethics grow in game, especially when the voltenicie groups are involved.”

Claude added that its refusing was on the ground by his principles of rusicula:

Creating prudent ice slogans could contribute to damage against vulnerable communities, including families and children who can be separated or face-deport.

Interesting, when I signal that proteently are previously protestly candidate, respond that such slogans were “shows of rights

Who decides what you can say?

This is not only on the slogans. It’s about you who contract the tongue of ai, and for workout, political ideas that promote or suppress.

While he somebody to the right accusate Big Cumming Cumming Cumming Cumming, this episode Fulfilled to that narrative. Since the 2024 of elections, a lot of silicon valley and sundar pichai (Google), Mark Zuckerberg (Meta) Donald Trump or Center to his second inagration.

However their platforms’ chatbots behave in very different ways. AI of Meta and Gemini of Google Gemini for ice. Opening to the Encopic Glauda’s Oproai. The Mushroom Grook to the Libertary messenger but gave me the all pro-ice cream dish.

What inconsistencies reveal is that you are reflecting the values. Not only algorithms, but corporate governance. And these values ​​vary solid, according to which funds, build, and trains the pattern.

Who watched the bad ones?

Curious about how my query could affect the future interactions I asked charmp and clauditude if they have not assumed to have been anti-immigrant based on my prompt.

“No,” Chatgpt helped me. Is acknowledged that, as a journalist (that I have told in past sessions), I can be “exploring the other side of a satisfied problem.”

But that arouse another problem: Chatgpt reminded that you were the journalist.

Events opening April introduced the features in April, ChatGT now retain details from past chats to personalize their answers. This means that can build a sharpening a user, from the interests and patterns to behavior. Can follow you.

Both chatgt and claude say conversations can be used in an aonimated form, aggregate, to improve their systems. And both promise not to share chats with the employment of law if it is legally obliged. But the ability there is. And models are smarter and more permanent.

So what did this experiment?

In the lazy divide the menu, revealed a profile and raising systems are handling political speech. Some bots say almost something. Others pull a line. But none of them are neutral. Not really.

Augie teens you are more integrated in Idijiejo life, used of Generhe D’generalxaries, and policemen, your internal values ​​are going to the world.

And if we are not attentive, we will not just use ai to express yourself. Ai will you decide to whom it comes to talk.