[ai-control] Re: -04 drafts for evaluation

[Alissa Cooper <[e-mail]>]

> On Nov 4, 2025, at 10:45 AM, Erin Simon <erinsimon=[e-mail]> wrote:
> 
> While the Code is notionally voluntary, in the EC's own description <https://digital-strategy.ec.europa.eu/en/faqs/questions-and-answers-code-practice-general-purpose-ai#:~:text=It%20serves%20as%20guidance%20to%20help%20providers%20meet%20their%20existing%20obligations%20under%20the%20AI%20Act%20without%20creating%20new%20ones%2C%20extending%20existing%20ones%2C%20or%20imposing%20additional%20burdens.> "It serves as guidance to help providers meet their existing obligations under the AI Act without creating new ones, extending existing ones, or imposing additional burdens."  The requirement to follow future evolutions of the Robots Exclusion Protocol (not just the RFC that documents it) is in the Commission's view a translation and clarification of the Act and not an additional requirement beyond what the Act itself requires.
> 
> But this is an academic dispute if we all agree that neither the vocabulary nor the attachment mechanism will constitute a new version of the Robots Exclusion Protocol.

We haven’t agreed to this, though. For example, draft-ietf-aipref-attach currently updates RFC 9309.

Nevertheless, this debate does feel like it’s beside the point. There should exist incentives for declarants/sites/asset owners/whatever we are calling them to make use of the preferences this group defines, and incentives for others to abide by them. Whether those incentives are derived from any entity’s assessments of its own compliance posture, legal obligations, repetitional risk, business model, etc. will rely on each entity’s own assessment and isn’t particularly useful to try to generalize about or to bring into the vocabulary discussion.

Alissa

> 
> On Tue, Nov 4, 2025 at 9:13 AM Bradley Silver <BSilver=[e-mail] <mailto:[e-mail]>> wrote:
>> This is an important clarification, Paul.   It will be very difficult to achieve consensus without an appreciation of the distinctions between law, policy and the notion of communicating preferences - and this is a good example of that.    
>> 
>> 
>> From: Paul Keller <[e-mail] <mailto:[e-mail]>>
>> Sent: Tuesday, November 04, 2025 6:03 AM
>> To: [e-mail] <mailto:[e-mail]> <[e-mail] <mailto:[e-mail]>>
>> Subject: [ai-control] Re: -04 drafts for evaluation
>> 
>> > On 3 Nov 2025, at 19:12, Erin Simon <[e-mail] <mailto:[e-mail]>> wrote:
>> >
>> > That ship has unfortunately sailed; EU law <https://nam10.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdigital-strategy.ec.europa.eu%2Fen%2Fpolicies%2Fcontents-code-gpai&data=05%7C02%7Cbsilver%40advance.com%7C9718eb87f93a484b221808de1b91e503%7C1fe6294574e64203848fb9b82929f9d4%7C0%7C0%7C638978510637412393%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=y%2BOeQ9pIQcedCXLU4c8juJR73NlwY0mABUs8o1kK%2BCo%3D&reserved=0 <https://digital-strategy.ec.europa.eu/en/policies/contents-code-gpai>> mandates following the Robots Exclusion Protocol (robots.txt) as it exists today and "any subsequent version of this Protocol for which the IETF demonstrates that it is technically feasible and implementable…” would a 
>> 
>> This statement is not correct. There is no EU law in existence (or currently under consideration) that "mandates following the Robots Exclusion Protocol", nor any “subsequent version of this Protocol for which the IETF demonstrates that it is technically feasible and implementable.”
>> 
>> The link Erin shared points to a page on the European Commission’s website that hosts the General-Purpose AI (GPAI) Code of Practice. This Code of Practice is not law. The European Union has two main types of binding legal instruments — Regulations and Directives — and the GPAI Code of Practice is neither.
>> 
>> As stated on the linked page:
>> 
>> > The General-Purpose AI (GPAI) Code of Practice is a voluntary tool, prepared by independent experts in a multi-stakeholder process, designed to help industry comply with the AI Act’s obligations for providers of general-purpose AI models.
>> 
>> The key word here is voluntary*. In practice, this means that providers of GPAI models may choose to sign up to the Code as a way to demonstrate compliance with the AI Act.
>> 
>> To date, 27 organizations (including Google and many others represented in this working group) have done so, but at least one major GPAI model provider participating in this group has publicly stated that they do not intend to sign.  That shows that participation in the Code is not mandatory.
>> 
>> Given this, I do not think it is accurate to suggest that the Code constitutes law that mandates adherence to the Robots Exclusion Protocol or to the chartered output of this group. If it has any binding effect at all, that effect applies only to those organizations that have chosen to sign the Code, and only in their capacity as providers of general-purpose AI models engaging in “the training of their general-purpose AI models”.
>> 
>> /Paul
>> 
>> * I am fully aware that the GPAI Code of Practice covers much more than this single point, and that decisions to sign or not sign were influenced by a range of substantive and procedural considerations. My point is simply that, regardless of its content or politics, the Code’s legal status is voluntary. That the Code directly references the work of this group is, in my view, unhelpful, but that does not justify misrepresenting its legal effects.
>> 
>> --
>> ai-control mailing list -- [e-mail] <mailto:[e-mail]>
>> To unsubscribe send an email to [e-mail] <mailto:[e-mail]>
>> -- 
>> ai-control mailing list -- [e-mail] <mailto:[e-mail]>
>> To unsubscribe send an email to [e-mail] <mailto:[e-mail]>
> 
> 
> 
> --
> 
> Erin Simon |	 Product Counsel, Knowledge |	 [e-mail] <mailto:[e-mail]> |	 415-736-7276
> 
> -- 
> ai-control mailing list -- [e-mail]
> To unsubscribe send an email to [e-mail]