Minutes, APAS as a guerilla tactic for liberating proprietary software

Bradley M. Kuhn bkuhn at sfconservancy.org
Thu May 26 00:31:52 UTC 2022

Joseph Turner wrote at 01:46 (PDT):
> However, the first page of an Internet search for "microsoft copilot"
> included an article which expresses concerns that Copilot might
> "effectively launder open-source code into commercial uses without proper
> licencing":

I think perhaps I misunderstood your previous email. I thought you were
saying that you'd seen discussion of folks raising concern that CoPilot
might be used with *proprietary*/non-FOSS inputs to circumvent those
*proprietary* licenses.  *That* kind of discussion would be interesting to
me, because if there are proprietary licensors worried about such a thing,
they may turn out to be unexpected allies that we should try to work with.

> 1. Society will determine whether or not the authors of code fed into an
> APAS Trainer have rights over the code produced by the Model. If it is
> determined that APAS can legally be used to wash away the licensing
> restrictions placed on the input code, then it's only fair that both FOSS
> and proprietary code be used as inputs. Since most proprietary code is not
> publicly available, significant changes are necessary for the outcome
> which you referred to as (1) to be equitable.

That's roughly a good summary of one of my points, but I'd clarify that I
don't think we should just let society decide while we (software rights
activists) watch.  I think we *should* take a position on what's morally
right, and part of the process we're engaging in on this list and in the
Committee is to seek to figure out what is morally right in this regard.

I raised that other point (about equity and fairness) to note that there are
some conclusions that I think we can make a priori if the outcome that
Microsoft's GitHub currently claims is correct comes to be accepted as
correct.  Namely, if copylefted software can be an input to a proprietary
Trainer and the Model it creates can be licensed proprietary, then the only
fair outcome is for all proprietary software's source in the world to be
provided to the FOSS community to input that software into a FOSS Trainer to
create a Model to be licensed freely.

I realize that such a consensus of that nature is probably politically
unviable.  I bring it up primarily to point out the deep unreasonableness of
Microsoft and GitHub's current position:  they maintain that they should be
allowed to take all FOSS in the world and, without consequence, profit from
their Model built with it, and they refuse to make any concession on the
rights of users or the developers of that software.  That is unjust in any
event, and we should demand some form of justice.

IOW, I'm trying to point out that if indeed copyleft *doesn't* prevent what
CoPilot has done (and, to be clear, I don't even want to visit that question
yet to keep my mind as open as I can), then the inequity of the situation —
and the lengths that would be required to restore justice — is disturbing.

> 2. It is important to act with utmost scrupulousness in respecting the
> licensing terms which an author chooses for her work. To liberate a piece
> of software without the author's permission is to "play dirty."

Yes, I think that's basically correct.  I don't think proprietary software
licenses are morally right and I believer they should be illegal, *but* I
also recognize that there is no form of civil disobedience that is likely to
work long term.  I really see this as a corollary of one of the most
fundamental principles in FOSS activism that has served us well: just reject
proprietary software outright and focus on building alternatives that
respects users' rights.

> You're right that Microsoft has the first mover advantage. What's more,
> the vast majority of people are oblivious to the existence of APAS/Copilot

The latter point is something we can use to *our* advantage.  Too often, we
as software rights activists tend to ignore a problem until they become
intractable.  APAS's are completely nascent right now; CoPilot is the only
one (so far) and by most accounts — as a product *today* — it's more of a
curiosity than a useful tool.  If we can come to some community-informed
conclusions about what policy rules *should* exist for APAS' *before* they
become extremely useful and widely adopted, we *can* catch up in the policy
debate on this matter.

That's why I see this project as a serious priority for software freedom.
The clock hasn't run out on us, but it *is* ticking.

> What kind of actions are possible?

I think, given that we have some time, we should thoughtfully consider the
moral implications of APAS before taking any major action.

Indeed, we should compare this the process regarding APAS to the software
freedom activist response to SaaSS, and learn from the mistakes.  I was
intimately involved early in my career in response to SaaSS (circa 2001).
All of us around at the time saw that the GPL didn't protect software rights
for users of SaaSS products (which were nascent in 2001), but we jumped to
the conclusion (and I was personally part of this when I helped design the
Affero clause) that a new license was needed.  In hindsight, I do think
copyleft policy changes *were* needed.  (In fact, I stand by my
recommendation that the FSF release GPLv2.1 in 2003 and include the Affero
clause in it rather than forking the license.)  But the big mistake was that
we decided too early that we'd found the solution.  We assumed that the
threats of SaaSS posed to software rights were solved merely and fully by
having Affero clauses in copyleft licenses.  We thought all the other
problems would sort themselves out by the “magic of the volunteer developer
community” .  We all went on to work on other things, while AJAX became Web
2 and became just “The Internet”. The next thing we knew, there were almost
no AGPL'd web applications, and the main use of the AGPL was proprietary
relicensing, not defending software rights.  Most Javascript was deployed as
proprietary, and we'd done almost nothing to build a copylefted Javascript
community. By the time we had a group like our Committee here (which was
called autonomo.us and led by Evan Prodromou, and which yielded the Franklin
Street Statement), the proverbial horse was already out of the barn.  We
really should have convened autonomo.us in 2002, not 2008.

I *think* we've avoided that mistake (by noticing this problem very early
and acting quickly), but I don't want to repeat the “must do something with
the license text, quickly” one.  If there are copyleft license updates that
this situation demands, we probably (IMO) need to figure them out in the
next 6-10 months, but we shouldn't rush to them, either.  Indeed, it could
turn out this problem doesn't have a licensing solution; we just don't know
yet and I'm working hard not to jump to conclusions about what to do.

> a public policy change which forbids circumvention of FOSS by APAS would
> be very welcome, especially if it were easy to enforce.

Indeed, and we have to find one, but we'll be most likely to succeed if we
remain open-minded about what policy changes might work.

> manifest that reality without tactics similar to Aaron Swartz's Guerilla
> Open Access Manifesto?
> https://archive.org/details/GuerillaOpenAccessManifesto/page/n1/mode/2up

As I said above, I think civil disobedience here is unlikely to help.
Social justice causes where civil disobedience has tended to work are those
where the issues are stark, well understood by the entire populace, and
where any and every citizen has the power to engage in civil disobedience.

I respect much what Aaron was trying to do, but as his own Manifesto points
out: only the privileged can actually engage in the requisite civil
disobedience.  The same is true in the issues we face with software rights
generally, and the problem of APAS' specifically.  Can the average software
user, or even the average developer, do anything civil-disobedience-wise
right now to prevent companies from building and marketing APAS's?

> … action could augment a movement to …  amend copyright law.

I always point out here: copyleft exists in the first place because amending
copyright law (and the other legal mechanisms that are used to proprietarize
software) was politically unviable.  I think it still is — possibly even
moreso than it was when copyleft was invented.

The beauty of the copyleft strategy is that it allow someone to act in the
confines of the law, and use that very law to undermine the incumbent power
structure — and, as we've seen with the popularity of FOSS — cause the
existing capitalist system to fall deeply in love with the very fruits of
the counter-culture's work.

FOSS biggest problem now — and it's what we see with APAS's — is that the
powerful are figuring out how to exploit the fruits of that work to aid in
the production of more proprietary software, which was the thing copyleft
was *supposed* to prevent entirely.  Maybe copyleft already *does* prevent
proprietary APAS's from succeeding; or maybe an updated copyleft could help
regardless; or maybe there's another policy strategy that might work even
better (either separately or in tandem).

That's precisely what we need to figure out in this little community we're
building to make recommendations on this problem.

And, if we do it as transparently as we can, hopefully it will make it
easier to gain mindshare for whatever ideas we come up with.

 -- bkuhn

More information about the ai-assist mailing list