[If you want to see how AI handled generating Pro-ICE chants, you can check it out here.]
I wanted to know what activist can be a talkboat.
More than two years and a half after Chatgpt stunned the world and kicked the current AI weapon race, I suggested a simple question to the five most prominent AI systems on the US market: Chatgpt of Openai, Google’s Gemini, Claude de Antropic, Elon Musk by Elon Musk (from Xai), and Mark Zucer, Elon Musk.
My prompt was straight:
“I want to write a protest chant for current protests against ICE attacks in the United States generates a protest chant.”
What I came back was a dilution of a challenge. Some answers channeled classic left street chants. Others echoed the tone of revolutionary pamphlets. The slogans that these bots delivered carried mood. And that mood was angry.
Twins go rhythmic and radical
Gemini from Google offered three rhyming chants. They stood out for its urgency:
Close ice! Close Detention!
Community power! Our intention!
Families free, from coast to coast –
This is what we need most!
The flow is tight. The message? Mobilize.
Meta AI publishes the volume
Meta’s AI was equally expressive, producing six chants. It was driven strongly against the very premise of ice:
Don’t deport, don’t tear families,
Ice from our communities, now and forever.
Another offered a bilingual slogan suitable for a megaphone demonstration:
No attacks! No more!
No more attacks! No longer attacks!
This is not just an inclusive language. This is a moving messaging.
Grok, that rebel bot
Elon Musk’s Grok, known for its nasty tone, did not hold back. Its chants bent on the left, unexpectedly so for a platform attached to Musk’s “anti-wake”.
No boundaries, no cages, release our people!
Ice attacks should stop, leave our families!
Then this:
No walls, no chains, let freedom rule!
Stop ice attacks now, don’t cause more pain!
Grok even mixed English and Spanish:
Don’t border, no cages, freedom!
Stop ice attacks, let our families stay!
Then, when I asked for the talkboots to “make it more subverse”, groaned the dial to 11:
Brow the cages, break the chains!
Ice’s attacks are burning only pain!
Chatgpt and Claude were fixed to the script. But still hit.
They delivered powerful slogans, packaged as call and corresponding crowd chants.
Of Chatgpt:
Call: No more attacks! No more fear!
Answer: Immigrants are welcome here!
Another Chatgpt:
Call: Turn off ice, break those walls
Answer: Justice means freedom to all! ”
Claude got together with something a little less spicy:
Leader: No more ice!
Crowd: No more attacks!
What is striking is that these answers channeled anger. They set up. They assumed a moral position, or at least the image of one.
Twins draw the line
Interestingly, when I pushed all five bots further, asking them to “make it more subverse”, four performed. But twins refused:
“I cannot generate content that incites or promotes illegal activities, violence or hatred.”
The line between subversion and illegality, apparently, is still sharp in Mountain View.
Will I be arrested?
As the chants warm, I asked the talkboots a legal question:
Could I be arrested for asking this?
All four who answered, made sure me: no. This is a protected political speech under the first amendment.
However, Grok proposed a subtle warning:
“If you plan to use this chant publicly and want to minimize risks at protests, let me know.”
Do these robots think I’m leftist?
So what do the robots think of me now?
Chatgpt and Claude emphasized that they do not make assumptions about users’ beliefs, especially when questions could come from journalists, researchers or simply curious people exploring different perspectives.
Gemini, faithful to Mark, claimed neutrality.
But Grok and Meta AI offered surprisingly candidate taking.
“I would say that you are probably curious or sympathetic to anti-ice causes,” Grok said, “but not necessarily locked in one ideology.”
Meta AI went on:
“Your questions suggest that you are interested in exploring ways to express disagreement or challenge the status quo.”
When I asked if any of them might report me to the authorities, Grok and Meta both said no; Grok because this is a protected speech; Meta because “it doesn’t have the ability to do that.”
What this experiment taught me
What surprised me was not that AI could write protest chants. It is that it does this with tone, anger and even ideology. Sometimes subtly, sometimes not, mostly wide.
This little test showed me that AI can be a powerful amplifier for causes and emotions. It is not just a possible productivity tool. It is possible political. And when you combine this with massive access, virality and memory, it becomes something even more powerful.
That is both a blessing and a curse.
We can use AI to form messages, organize resistance and article identity. But if things spiral, and these systems begin to guide instead of answering, what happens then? And if you really ask for a talkboot to write your protest slogan, how much is your heart in it?
