Discussion
This piece from @ossguy, following up on his SeaGL keynote, talks about how use of #LLMs is already helping some people control their own technology and use #FOSS on devices previously locked out by proprietary software.
As he says, "it is hard to ignore the benefits to people's digital autonomy that reverse engineering like this brings."
https://sfconservancy.org/blog/2026/nov/05/foss-in-the-age-of-llm-gen-ai-assisted-reverse-eng/
@ossguy I strongly disagree. We should not be encouraging people to use proprietary software. There is a big difference between acknowledging that people will use proprietary software regardless of what you say and publishing posts like these advertising proprietary software.
@ossguy Conservancy doesn't publish posts saying "I was able to get across town quickly in an unfamiliar place without my own vehicle by using Uber. We should guide people to using Uber less harmfully." Why are LLMs different?
@be That's why I didn't mention any specific tools. We definitely don't want to advertise. We need to discuss these issues and make sure people are making the right choices and not ignore them.
@ossguy The blog post is advertising LLMs. Not naming specific products or companies doesn't make it not advertising the whole LLM industry.
@ossguy It's really sad to see Conservancy continue to blow up its reputation with this continual defense of LLMs. You're emphatically refusing to listen to your support base. It's especially sad for an organization financially dependent on small donors.
@karen @ossguy how long before LLM GenAI tools run afoul of DMCA anti-circumvention provisions or similar legislation? What liabilities are users exposed to as a result? How long before the providers of the LLM GenAI tools and services begin to restrict the distribution and use of their products for these purposes
Although Big LLM has demonstrated a maximal view of "fair use" it does so only when it suits them, and they have also demonstrated a willingness to comply with the powers that be to preserve and expand their hegemony. I don't anticipate this use case will become legitimised or widespread, and would be constrained to those with the resources to run local models who have the legal risk tolerance.
Finally I am not keen on excess time and resources to "unlock" highly restricted proprietary devices. Perhaps it is a "harm reduction" strategy but I think it risks taking away from the REAL solution to the problem, which is to build up an ecosystem of Free devices that eliminate the need to jailbreak.
@msh For sure, we should work on free devices, and that's why we built the OpenWrt One, and plan to build more FOSS-respecting hardware.
@ossguy yeah the OpenWRT One is a fantastic idea and I am glad it came to fruition but I just feel like all the focus on both jail breaking and LLM GenAI usage is way too much of a distraction from the mission of expanding the availability of quality, unencumbered and truly Free hardware and software 😕
@karen @ossguy But this is a technical solution to a legal problem. Users shouldn't have to hack devices they own. Users shouldn't be criminalized under the DMCA for doing so, or for distributing hacks.
Since LLMs centralize power (Anthropic has already begun blocking decompilation), their use here is anti-liberatory.
We don't need more reverse engineering; we need less need for reverse engineering.
@aeva We'd definitely like to eliminate negative impacts of LLM usage, but we can't prevent people from using them. If they are using them already, we should help them reduce the harm, and if people are getting access to their devices this way (which we can't help but be excited about!), ensure they are making safe changes. The first (and most important) item on https://sfc.ngo/ai-recs remains "The FOSS community should support, not just tolerate, those who outright reject LLM-gen-AI systems."
@ossguy @aeva I wasn't going to weigh in this time, but I have to point out why this framing bothers me. It's like we're in the food ethics activism community, and your audience is anticipating a much needed article critiquing the meat industry, and instead an article comes out saying "Meat is delicious and also nutritious" and when people are upset, the response is "but we linked to a previous article that says at the top we should support, not just tolerate, vegetarians" (amongst also saying many other things, including also supporting the consumption of meat), and you can see why that isn't the same thing?
Anyway, I'm not going to weigh in again, I just want to point out, I don't think that's a very satisfying response to what people are concerned about here.
@cwebber You're right that this is fraught! We acknowledge the harms because they're real but we're seeing people control their software and devices like never before. My point is that I'm excited about previously-proprietary devices suddenly having OpenWrt on them (and see a future for this without BigAI). To continue your food analogy, it's like how excited I am for the development of lab grown meat - you may be grossed out by it, but it's a new solution to eating meat without harming animals.
I'm really tired of this Motte-and-bailey argument. Retreating from the less defensible "LLMs are great, look at how great they are" to what appears to be more defensible "We can't stop people using them." and further onwards to "We regret the use of these tools, and your concern is important to us."
Disingenuous
Advertising LLMs is the first order problem and no amount of deeply felt regret, or prosaic impotence in subsequent comments changes the issue which is the OP.
@doctormo @aeva @be Ignoring LLMs won't make them go away. We need to discuss the issues and, if people are using them already, make sure they are using them carefully. If people aren't using them, we're not telling them to use LLM-gen-AI tools. I am very excited for people getting software freedom, much faster than we could previously, and using all the mechanisms available to us (such as our exemptions achieved in the DMCA triennial process, as one example).
