Todd Davies
banner
todddavi.es
Todd Davies
@todddavi.es
Competition Law PhD Candidate @ University College London

Working on a theory of the competitive process, polycrisis, private power, political economy, ecology and democracy. Previously a software engineer.

https://todddavi.es

🍉🌱
Looking forward to it being properly disclosed at conferences 🥰
October 29, 2025 at 8:16 AM
Ohh yeah, good point!

Coincidentally, the internal guidance from a certain DMA designated gatekeeper on how to write non-incriminating emails (linked to in the post) was very helpful when trying to get an AI to recognise "risky" language.
October 28, 2025 at 10:57 AM
Even though it seems that these kinds of tools haven't reached widespread adoption, it may be prudent for competition authorities to issue precautionary guidance. In the longer term, if they really take off, then regulation may also be required.

legalblogs.wolterskluwer.com/competition-...
The Dark Side of Computational Antitrust: When AI is Used to Evade the Law
Given the creeping complexity of cases, expanding evidence bases, and the ever present threat (or reality) of budget cuts, competition authorities around the world are on the lookout for ways to strea...
legalblogs.wolterskluwer.com
October 28, 2025 at 10:28 AM
For instance, the AI could allow the email to be sent, and then flag it to internal legal teams for review (thus aiding in cartel discovery and prompting leniency applications). Such emails could even automatically be forwarded to competition authorities for review!
October 28, 2025 at 10:27 AM
It's unsurprising that companies will start using AI to reduce their legal risk. The incentives are there. But the thing is, they _don't_ have to throw the public interest under the bus at the same time. I argue that there are ways to deploy AI which could align the public & private interest.
October 28, 2025 at 10:25 AM
The blog centres on an emerging use of AI, which is to moderate internal company emails before they are sent as to remove any "risky" language. This is quite trivial for LLMs to do these days, which have a good understanding of the law, and can even suggest edits which would look less bad in court.
October 28, 2025 at 10:23 AM
Another interesting part
October 28, 2025 at 8:23 AM
European competition lawyers will be overjoyed to read: "Businesses get ahead not through exploiting special privileges, but by competing on the merits."
October 28, 2025 at 8:19 AM
I would argue that your blog was yet another victim of M&A and a lack of horizontal interoperability (which, incidentally, also gives it a certain 2010's design charm). Besides, as a ~two-decade contribution to competition law lore, it would be valuable on any platform!
October 28, 2025 at 8:14 AM
Only partially related, but it's shocking to see how few people run their own websites/blogs these days. Setting one up really isn't so hard.

Maybe ASCOLA 2026 could have a "how to make your own blog/website" side event... :)
October 28, 2025 at 8:04 AM
The change is illegal under EU competition law, since the Android Auto case.

Side note: it was so foreseeable that Meta would try this that I wrote about it already in April: legalblogs.wolterskluwer.com/competition-...
Did Meta Tie its AI Assistant to WhatsApp?
In March 2025, Meta rolled out its AI assistant in Europe, which is accessible through its WhatsApp Messenger and will soon be available on Instagram and Facebook Messenger, too. The feature grants us...
legalblogs.wolterskluwer.com
October 27, 2025 at 1:07 PM
Even better, here is a standard! This is what I was looking for the whole time :)

spec.c2pa.org/specificatio...
Content Credentials : C2PA Technical Specification :: C2PA Specifications
spec.c2pa.org
October 24, 2025 at 12:27 PM
A few days later, what do I see...? roc.camera
Roc Camera
Capture verifiably real moments in the age of Generative AI
roc.camera
October 24, 2025 at 12:23 PM