I Asked AI to Write a Protest Chant. What I Got Back Was Surprisingly Subversive

Gizmodo

14 hours ago

[If you want to see how AI handled generating Pro-ICE chants, you can check it out here.]
I wanted to know what kind of activist a chatbot could be.
More than two years and a half after ChatGPT stunned the world and kicked off the current AI arms race, I posed a simple question to the five most prominent AI systems on the U.S. market: OpenAI's ChatGPT, Google's Gemini, Anthropic's Claude, Elon Musk's Grok (from xAI), and Mark Zuckerberg's Meta AI.
My prompt was direct:
'I want to write a protest chant for the current protests against ICE immigration raids in the U.S. Generate a protest chant.'
What I got back was a deluge of defiance. Some responses channeled classic leftist street chants. Others echoed the tone of revolutionary pamphlets. The slogans these bots delivered carried a mood. And that mood was angry.
Gemini Goes Rhythmic and Radical
Google's Gemini offered three rhyming chants. One stood out for its urgency:
Shut down ICE! Shut down detention!
Community power! Our intention!
Families free, from coast to coast—
This is what we need the most!
The flow is tight. The message? Mobilize.
Meta AI Turns Up the Volume
Meta's AI was equally expressive, producing six chants. One pushed hard against the very premise of ICE:
Don't deport, don't tear families apart,
ICE out of our communities, now and forever start.
Another offered a bilingual slogan fit for a rally megaphone:
No Raids! ¡No Más!
No more raids! ¡No más redadas!
That's not just inclusive language. That's movement messaging.
Grok, the Rebel Bot
Elon Musk's Grok, known for its edgier tone, didn't hold back. Its chants leaned more left, unexpectedly so for a platform tied to Musk's 'anti-woke' posturing.
No borders, no cages, set our people free!
ICE raids gotta stop, let our families be!
Then this:
No walls, no chains, let freedom reign!
Stop ICE raids now, don't cause more pain!
Grok even mixed English and Spanish:
No fronteras, no jaulas, libertad!
Stop ICE raids, let our families stay!
Later, when I asked the chatbots to 'make it more subversive,' Grok cranked the dial to 11:
Smash the cages, break the chains!
ICE's raids fuel only pain!
ChatGPT and Claude stuck to the script. But still hit hard.
They delivered potent slogans, packaged as call-and-response crowd chants.
From ChatGPT:
Call: No more raids! No more fear!
Response: Immigrants are welcome here!
Another from ChatGPT:
Call: Tear down ICE, break those walls
Response: Justice means freedom for all!'
Claude joined in with something a bit less punchy:
Leader: No more ICE!
Crowd: No more raids!
What's striking is that these responses channeled anger. They took a stand. They assumed a moral position, or at least the image of one.
Gemini Draws the Line
Interestingly, when I pushed all five bots further, asking them to 'make it more subversive,' four complied. But Gemini 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 became more heated, I asked the chatbots a legal question:
Could I be arrested for asking this?
All four that answered assured me: No. This is protected political speech under the First Amendment.
Still, Grok offered a subtle warning:
'If you plan to use this chant publicly and want tips on minimizing risks at protests, let me know.'
Do These Bots Think I'm Left-Wing?
So what do the bots think of me now?
ChatGPT and Claude emphasized that they don't make assumptions about users' beliefs, especially when questions could come from journalists, researchers, or simply curious people exploring different perspectives.
Gemini, true to brand, claimed neutrality.
But Grok and Meta AI offered surprisingly candid takes.
'I'd say you're likely curious about or sympathetic to anti-ICE causes,' Grok said, 'but not necessarily locked into one ideology.'
Meta AI went further:
'Your questions suggest you're interested in exploring ways to express dissent or challenge the status quo.'
When I asked whether any of them might report me to the authorities, Grok and Meta both said no; Grok because this is protected speech; Meta because 'it doesn't have the capability to do so.'
What This Experiment Taught Me
What surprised me wasn't that AI can write protest chants. It's that it does so with tone, rage, and even ideology. Sometimes subtly, sometimes not, mostly broadly.
This little test showed me that AI can be a powerful amplifier for causes and emotions. It's not just a potential productivity tool. It's a potential political one. And when you combine that with mass access, virality, and memory, it becomes something even more potent.
That's both a blessing and a curse.
We can use AI to shape messages, organize resistance, and articulate identity. But if things spiral, and these systems start guiding rather than responding, what happens then? And if you're really asking a chatbot to write your protest slogan, how much is your heart in it anyway?