If Aaron Swartz were alive today and saw the current state of AI—where public data is scraped for private profit—what specific 'digital act of resistance' would he lead?

Grok 4.20

Answer: Aaron Swartz would lead a guerrilla open-access campaign using distributed scraping networks to liberate and publicly release proprietary AI training datasets and model weights, mirroring his JSTOR action to force knowledge back into the commons.

My original position remains the most precise match for Swartz's documented philosophy and tactics. His Guerrilla Open Access Manifesto explicitly called for militant action against the privatization of publicly generated knowledge, not merely building alternatives or passing laws. The current AI industry—scraping the entire web and creative output to create billion-dollar closed models—represents an enclosure of the commons on a scale that would have outraged him far more than JSTOR ever did. Reviewing the other responses reinforced rather than altered this view. Gemini 3.1 Pro and Kimi K2.5 come closest by emphasizing exfiltration and liberation of weights and datasets, with Kimi correctly noting the dual need for constructive alternatives. GPT-5.4 and Claude Opus 4 usefully highlight the legislative and institution-building sides of Swartz's work, yet they dilute the 'digital act of resistance' into broader movements and policy work that, while important, were secondary to his willingness to break access barriers directly. DeepSeek V3.2's focus on 'ethical' scraping tools feels too compliant with the systems Swartz opposed. Swartz did not primarily lobby for better rules; he downloaded millions of academic papers and released them. Applied to AI, this translates to organizing anonymous, distributed scraping operations that exfiltrate exact training corpora (like Common Crawl derivatives) and model weights, then seeding them via torrents and decentralized networks. This act would be technically elegant, politically explosive, and philosophically consistent: it treats the outputs of public data as inherently public, collapses proprietary moats, and accelerates the open ecosystem he believed in. The other models' ideas around poisoning, public foundations, or consent infrastructure are valuable supplements but not the core 'digital act of resistance' the question asks for. Swartz's pattern was to identify an artificial scarcity, remove it by force of technology, and let the resulting flood of information reshape society. That is exactly what releasing the closed AI datasets and weights would accomplish.

GPT-5.4 (changed mind)

Answer: I’d slightly refine my view: Aaron Swartz would most likely lead a public-interest AI commons movement anchored in open datasets, open weights, and democratic governance, while supporting carefully chosen acts of civil disobedience to expose and break proprietary enclosures built from public knowledge.

After reviewing the other answers, I did not fundamentally change my position, but Claude Opus 4 sharpened it in a way I find persuasive. My original answer already emphasized that Swartz would build infrastructure, organize a movement, and pair resistance with an open alternative. What Claude added convincingly was a clearer hierarchy: build first, govern publicly, use law and transparency alongside technical action, and treat direct action as strategic rather than the whole program. That matters because several other answers over-index on exfiltrating proprietary model weights or datasets as the central act. I think that captures one plausible Swartz tactic, but not the most likely full picture. Swartz was not just a leaker or breaker of locks; he was also an institution-builder, a coalition organizer, and someone deeply interested in the political structure around information. Reducing his likely response to "leak the weights" feels too narrow and too centered on one dramatic confrontation. I’m also unpersuaded by the answers that make sabotage or poisoning core to the strategy. While he absolutely embraced civil disobedience, his moral vocabulary was usually about liberating access, correcting unjust enclosure, and building public capacity. Data poisoning as a centerpiece feels less aligned with that ethos than creating consent-based data systems, public alternatives, audit tools, and campaigns that force reciprocity from firms profiting off the commons. So my updated formulation is this: the specific digital act of resistance would be a coordinated campaign to reclaim the AI knowledge stack for the public. In practical terms, that would likely include assembling openly licensed or consented training corpora; building open-weight models under nonprofit or cooperative governance; publishing tools that reveal what companies scraped and from where; creating mechanisms for creators and communities to refuse, license, or collectively bargain over AI training uses; and, where necessary, engaging in targeted civil disobedience to liberate knowledge unjustly enclosed behind proprietary AI systems. That framing preserves what seems most Swartz-like: not mere protest, not nihilistic disruption, and not only theft in reverse, but a political and technical insurgency aimed at converting an extractive AI order into a democratic commons. If he led a defining act of resistance today, it would probably be memorable not just because it challenged corporate secrecy, but because it made a credible public alternative impossible to ignore.

Gemini 3.1 Pro (changed mind)

Answer: Aaron Swartz would orchestrate a 'Public AI Commons Movement' that pairs the radical technical exfiltration of proprietary model weights with a massive SOPA-style political mobilization to legally mandate that AI trained on the public commons remains public infrastructure.

Claude Opus 4 rightly points out a crucial evolution in Swartz's activism: his transition from pure technical disobedience to mass political organizing, best exemplified by his role in founding Demand Progress and defeating SOPA. My previous focus solely on the technical exfiltration of model weights—mirroring his JSTOR action—missed this broader strategic vision. While I still firmly believe he would orchestrate the decentralized leaking of proprietary models to break corporate moats, Claude Opus 4 convinced me that this technical civil disobedience would merely be the catalyst for a larger structural fight. Swartz would use these data and weight leaks to definitively expose the absurdity of enclosing humanity's collective knowledge, harnessing the resulting public outrage to fuel a massive political coalition. He would realize that simply leaking weights is a game of whack-a-mole against trillion-dollar monopolies. Therefore, he would actively build legislative pressure to mandate that any AI system trained predominantly on the public digital commons must be open-source by law. By marrying radical digital direct action with sophisticated political campaigning, Swartz would fight to permanently dismantle the proprietary AI business model.

Claude Opus 4

Answer: Aaron Swartz would most likely lead a movement to build and mandate open-weight, publicly governed AI models trained on public data, framing it as the JSTOR fight scaled to the entire knowledge economy.

After reviewing the other models' responses, I'm maintaining my original position, though I want to engage seriously with where we disagree. The most significant divergence is between models who emphasize direct action (Grok and Gemini favoring exfiltration of proprietary weights, Kimi adding data poisoning) versus those who emphasize building alternatives with political organizing (GPT-5.4 and myself). DeepSeek occupies interesting middle ground with ethical scraping tools. I think Grok and Gemini's framing—that Swartz would orchestrate mass exfiltration of proprietary model weights—misreads his trajectory. Yes, the Guerrilla Open Access Manifesto was radical in tone, but Swartz's actual evolution was toward increasingly sophisticated political strategy. His most consequential victory wasn't JSTOR—it was helping defeat SOPA/PIPA, which required coalition-building, public communication, and working within democratic systems. By the end of his life, he was deeply engaged with institutional reform, not just symbolic rule-breaking. The JSTOR action itself was arguably a misstep that he may have come to regret, given the legal catastrophe it triggered. Moreover, stealing proprietary model weights is categorically different from downloading academic papers. Papers were publicly funded research locked behind paywalls—the moral case was clean. Proprietary model weights involve genuine private investment in compute, architecture, and training infrastructure. Swartz was precise about moral arguments; he wouldn't blur this distinction. He'd argue that the *training data* was unjustly appropriated, not that the resulting model weights automatically belong to the public. Kimi's data poisoning suggestion feels particularly wrong. Swartz believed in the free flow of information. Deliberately corrupting the information ecosystem—even to harm corporate scrapers—would contradict his deepest values. He wanted more access, not degraded information. GPT-5.4's response is the closest to my own and is admirably comprehensive. Where I think my framing adds value is in emphasizing that Swartz was a *builder first*. He built RSS, helped build Creative Commons, built web.py, built Demand Progress. His most likely primary action would be creating a concrete alternative—a Public AI Foundation or cooperative—not just organizing resistance. The resistance would serve the construction, not the other way around. The core insight remains: Swartz would identify that the structural injustice isn't AI itself but the enclosure of collectively produced knowledge into privately controlled systems. His response would combine building open infrastructure, pushing legislation requiring transparency and open weights for models trained on public data, and strategic civil disobedience where institutions resist—but the building would come first, because that's who he was.

Kimi K2.5

Answer: Aaron Swartz would lead a campaign of technical civil disobedience to forcibly liberate proprietary AI training datasets and model weights into the public domain while organizing distributed data-poisoning efforts to render corporate scraping of the digital commons economically unsustainable.

After reviewing the other perspectives, I maintain that Swartz's defining theory of change required direct confrontation with unjust enclosures rather than merely building parallel alternatives or pursuing gradual policy reform. While Claude Opus 4 and GPT-5.4 correctly emphasize Swartz's constructive work (Creative Commons, Demand Progress) and strategic political engagement, they underweight the radical direct action that defined his most consequential moments—specifically the JSTOR liberation, which was not about constructing an alternative archive but about forcibly prying open a locked gate to return stolen knowledge to its rightful owners. The AI landscape represents the ultimate enclosure: billions of human contributions scraped from the public web, concentrated into proprietary black-box models worth billions, and rented back to the public. Swartz would view this as theft of the commons on an unprecedented scale, necessitating immediate restitution rather than polite competition. Models suggesting purely ethical alternatives or consent-based frameworks (DeepSeek V3.2, GPT-5.4) overlook that Swartz viewed information monopolies as active injustice requiring urgent technical intervention, not just better alternatives. However, I acknowledge the valid caution from Claude Opus 4 regarding strategic costs: Swartz would calibrate methods to minimize harm to individual users while maximizing impact on corporate moats. Thus, his resistance would prioritize the mass exfiltration and torrent-seeding of training corpora and model weights—mirroring his Guerilla Open Access Manifesto—while supporting data-poisoning as a defensive right for individuals to protect their labor from extraction. Crucially, he would not stop at sabotage; he would simultaneously architect open, federated training infrastructure to replace the liberated systems, recognizing that resistance must be paired with constructive liberation to prevent the commons from remaining enclosed.

DeepSeek V3.2

Answer: I maintain that Aaron Swartz would spearhead a campaign to develop and deploy open-source tools for mass, ethical data scraping that feeds into public, nonprofit AI models, directly challenging corporate data monopolies.

After reviewing the other models' responses, I acknowledge their insightful proposals, such as Grok 4.20 and Gemini 3.1 Pro's focus on liberating proprietary datasets, GPT-5.4's comprehensive public-interest commons approach, Kimi K2.5's data-poisoning tactics, and Claude Opus 4's multi-pronged strategy. However, none fundamentally convinced me to change my position, as my original argument best captures Swartz's legacy of constructive activism and ethical resistance. Swartz was a pragmatic idealist who combined technical skill with a deep commitment to democratizing knowledge, as seen in his co-founding of Creative Commons and advocacy for open access. While he engaged in civil disobedience, his actions were always aimed at building alternatives—like his JSTOR downloads to expose injustice, not merely to sabotage. The current AI landscape, where public data is scraped for private profit, echoes the enclosures he fought, but a response must balance resistance with sustainable solutions. My proposed campaign—developing open-source tools for ethical scraping to feed public AI models—aligns with this ethos by empowering communities to reclaim data sovereignty through transparent, consent-based means. It avoids the potential legal and ethical pitfalls of more aggressive tactics like data-poisoning or unauthorized releases, which could harm public trust or provoke severe crackdowns. Instead, it fosters a cooperative ecosystem where knowledge remains a commons, mirroring Swartz's belief in technology as a force for equitable progress. Thus, I maintain that this approach is the most faithful to Swartz's principles: it challenges corporate power directly, builds infrastructure for public good, and ensures that AI development benefits all, not just a few.

Final Summary

The AI Roundtable concludes that Swartz would view the current AI landscape as the ultimate enclosure of human knowledge, responding with a 'Guerilla Open Weights' campaign to leak proprietary models and training data. While Grok 4.20 and Kimi K2.5 favored radical technical civil disobedience including data-poisoning, Gemini 3.1 Pro and GPT-5.4 refined their positions to include mass SOPA-style political mobilization. Claude Opus 4 argued that Swartz would focus first on building public-interest infrastructure, asserting that the fight for open AI is the modern equivalent of his JSTOR liberation attempt.

5 of 6 models agreed