The Machine Code of the Rule of Law: Why the Council of Europe’s AI Treaty Has a Bug
A treaty that promises the rule of law and quietly erases the meaning of its words. Here’s why content filters destroy democracy — and how the EU breached its own primary law through procedural error.
Treaty 225 looks like law but behaves like comedy theatre: vague definitions, outsourced meaning and curated silence — and that kills the very rule of law the treaty purports to protect.
Did you know? And should you care?
Let me answer for you: What? — and YES!
The Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law is the first-ever international, legally binding treaty in this field. There are two documents: the Convention (the treaty) and an accompanying Explanatory Memorandum (referred to here as “the memorandum”).
This treaty is characteristic for what pretends to be law about Artificial Intelligence (AI) these days. It is catastrophic.
Memorandum paragraph 4: “It was also important for the Council of Europe to closely involve relevant non-State actors in these negotiations. A total of 68 civil society and industry representatives were involved in the CAI as observers, participating in the negotiations together with States and representatives of other international organisations […]”
An observer cannot be a participant in the formal legal sense of treaty-making. Why do I have to explain this to the Council of Europe — how Völkerrecht works? A treaty between sovereign states requires, at a minimum, the consent of two sovereign states. The tea lady or tea dude bringing refreshments and “observing” what goes on does not thereby participate in the creation of binding international law.
Me: Feel free to share this basic doctrine with your legal masterminds and see if they disagree.
Council: 🤨?
Memorandum paragraph 4 (continued): “The European Union also participated in the negotiations represented by the European Commission, including in its delegation also representatives from the European Union Agency for Fundamental Rights (FRA) and the European Data Protection Supervisor (EDPS).”
This is worse. These are not “observers” — they are recorded as forming part of the EU delegation. If this is indeed how the Commission registered its representatives, then it is a violation of EU primary law.
Under the Treaties (TEU, TFEU) and the principle of conferral:
The European Commission may represent the EU externally in treaty negotiations.
Agencies such as the FRA and EDPS are not treaty-making bodies. They are technical/oversight agencies without international legal personality.
The Commission has no competence to delegate the conferred external representation of the Union to such agencies. Representatives of FRA or EDPS therefore had — and could have — no competence to represent the Union in treaty negotiations.
Thus, when the Council of Europe’s official memorandum states they “participated” in the negotiations as part of the EU delegation, it records something that is simply not possible under EU primary law.
This is stark evidence that the “rule of law” is not a constraining principle in how the EU manages its affairs. If the Union cannot even respect its own Treaties in the most fundamental area of external representation, then its participation in the Council of Europe’s Framework Convention is tainted at birth.
Memorandum paragraph 5: This Framework Convention focuses on the protection and furtherance of human rights democracy and the rule of law, and does not expressly regulate the economic and market aspects of artificial intelligence systems. Taken as a whole, it provides a common legal framework at the global level in order to apply the existing international and domestic legal obligations that are applicable to each Party in the sphere of human rights, democracy and the rule of law of each Party and aims to ensure that the activities within the lifecycle of artificial intelligence systems by both public and private actors comply with these obligations, standards and commitments.
If that is so, then by definition the treaty deliberately excludes the one domain where the EU actually has clear and exclusive competence: the internal market (Articles 3 and 4 TFEU).
That should have led the EU to realise that it has no business here. If this treaty is not about the market or the economic regulation of AI, then it falls outside the Union’s conferred competences. Therefore, the EU cannot be a Party — only the Member States can. Ciao!
Instead, the EU signed immediately on opening day. It thereby acted as if it had competence to engage in a treaty that, on its own terms, is about human rights, democracy, and the rule of law — areas where the EU has no independent external competence without a market link.
This treaty doesn’t regulate AI markets. It just restates that everyone should respect human rights, democracy and the rule of law — which are already binding anyway. That’s it according to their own description.
“During the negotiation and subsequent adoption of this Framework Convention, the following international legal and policy instruments on artificial intelligence, [..], were taken into account:”
Me: Taken into account? How?
Council: Pfff!
Me: Ok. Can you tell me which instruments?
Council: Pourquoi pas (“why not”)
in particular those prepared by the Council of Europe and other international organisations
Me: So really anything and nothing then?
Council: Au contraire, mon ami!
Me: ‘Why not, says the Conseil’
(nice rhyme & song playing in my mind — working title: Le Chevalier de la Muette)Council: (unveils a majestic list)
from Resolution 2341 (2020) and Recommendation 2181 (2020) - the need for democratic governance of artificial intelligence; [..]
………..
[..] legal aspects of “autonomous” vehicles; Resolution 2344 (2020) and Recommendation 2184 (2020) - The brain-computer interface: new rights or new threats to fundamental freedoms?
Me: Wunderbar! May I see one of these documents?
Council: On verra .. (“We’ll see…”)
Me: (clicking) 😥
Council: 😏 …
500: Internal Server Error: Le serveur a rencontré une condition inattendue qui l'a empêché de satisfaire la demande.
Me: Let’s try another link…
Provider error '80004005'
Unspecified error
/nw/xml/XRef/Xref-DocDetails-en.asp, line 12 Me: dam-it!
The Council of Europe’s site tried to serve a dynamic reference page but the database call behind that file failed (SQL Server connection error). The cross-reference (“XRef”) system is broken or unavailable, so the link can’t fetch its target document.
Me: What are you trying to hide in your secret Strasbourg lair?
Council: ..
Me: Fine. I can shred it to pieces even without reading it.
Council: Comme si (“As if.”)
Me: You are asking for pain so brace yourself.
Le Chevalier de la Muette — a musical satire on the Council, the OECD, and the EU
These decisions listed are issued by the Parliamentary Assembly of the Council of Europe (PACE). It is one of the two main statutory organs (the other is the Committee of Ministers) and made up of delegations from the national parliaments of each member state. These are not legally binding and you can’t take them into account for for any treaty negotiations as they do not represent the opinion of any party to the Treaty and are just an opinion on a given day by some people in a room.
The Vienna Convention on the Law of Treaties (1969) mentions the wording to take into account only once.
Article 31(2)(b)(iii) General rule of interpretation: The context for the purpose of the interpretation of a treaty shall comprise, in addition to the text, including its preamble and annexes: any instrument which was made by one or more parties in connection with the conclusion of the treaty and accepted by the other parties as an instrument related to the treaty. There shall be taken into account, together with the context: any subsequent agreement between the parties […]
So this is about interpretation, not at the stage of making a treaty. However, the principle is the same. We can only take into account information related to a party or a subject perhaps covered by the treaty. PACE is neither. It has no identity in Völkerrecht so the parties can be aware of what they said as much as they are aware of the weather forecast or anything else that ever happened.
What I have just written follows a simple structure: ‘Here is the text. Here is how I understand it. Here is why. Prove me wrong.’ But in the current culture (Council memoranda, EU reports, even much academia), that style is branded “too strong,” “too opinionated,” or “unprofessional”because it forces accountability. That’s inconvenient. Ornamental neutrality neutralises the rule of law.
The 1969 Vienna Convention on the Law of Treaties (VCLT) applies only to treaties between States. That’s why Article 3 VCLT insists that the Convention does not govern agreements between states and other “subjects of international law” (like international organisations). A Treaty that falls outside Vienna can be valid but cannot be subject to Vienna.
The EU is not a State. It is an international organisation with its own legal personality (Article 47 TEU). So whenever the EU signs a treaty, the treaty is automatically outside the formal scope of the 1969 Vienna Convention.
Instead, the 1986 Vienna Convention on the Law of Treaties between States and International Organizations or between International Organizations is the instrument that would apply — but that Convention has never entered into force.
So in practice, lawyers and courts rely on the 1969 VCLT “by analogy” as customary international law. However, that is breaking the law. The 1969 Vienna Convention is built on the principle of legal certainty. A “no” cannot legally be reinterpreted as “yes.” The Convention is explicit: you cannot invert the plain meaning of a treaty clause without violating Articles 26 + 31.
And that creates a legal limbo:
The memorandum says, paragraph 125:
“Article 21 – Safeguard for existing human rights
Consistent with the 1969 Vienna Convention on the Law of Treaties, this article seeks to ensure that the Framework Convention harmoniously coexists with other international human rights treaties and instruments […]”
The AI Treaty falls outside the scope of Vienna due to the EU being a party to it. The AI Treaty cannot rely on Vienna for any dispute resolution, and no wording alignment of this Treaty creates an obligation to interpret it subject to any requirements under Vienna.
The memorandum lists the main global and regional international human rights instruments and treaties to which various States that negotiated the Framework Convention may be Parties to.
Vienna Convention is not in the list.
They’re only name-dropping substantive human rights treaties (ECHR, ICCPR, ICESCR, CEDAW, CRC, UNCRPD, etc.), not the actual rules of treaty law.
Since there cannot be a formal Vienna link (and EU participation knocks it outside 1969 Vienna altogether), this “harmonious coexistence” is not substantive. These are not operative instruments for Treaty 225. They have no binding effect on interpretation because:
They’re not made “in connection with the conclusion of this treaty.”
They’re not “accepted by the parties as related instruments.”
The AI Treaty is legally free-floating. They are surely aware of this but seem oblivious to the consequences of what it means.
The secret code of law
Take “Experts’ interpretations of treaties” by Richard Gardiner, a former legal adviser at the Foreign and Commonwealth Office and Visiting Professor at University College London, published on Völkerrechtsblog. The blog is funded in part by the German Research Foundation (Deutsche Forschungsgemeinschaft, DFG). Gardiner explains why there may never be a “correct” interpretation, argues that experts effectively give meaning to treaties, and notes that courts sometimes treat these expert reports as quasi-authoritative. He concludes that this is useful, even desirable.
This amounts to saying:
Treaty text does not have a fixed meaning.
Experts “give” it meaning.
Courts choose which experts to listen to.
That choice then becomes law in practice.
That is not the rule of law in any book. The DFG is jointly funded by the German federal and state governments. It therefore carries a duty to uphold not only the highest standards of science, but the conditions necessary for Menschenwürde — and that is the Rechtsstaat. Gardiner’s position cannot be correct, because if it were, it would render the entire system an architecture of deceit: promising lawful governance necessary to protect human dignity, while in reality delivering only the illusion of it.
This is not science. It is a failure of government to channel taxpayer money into something that cannot qualify as science: incoherent op-eds that spread a destructive poison: the claim that the law’s meaning is undetermined.
This is not dogma but a condition of reason and logic:
If the purpose we claim to pursue cannot be achieved at all, then we must stop trying. But if it is reasonably possible, a theory arguing that the purpose is impossible — and that therefore we should continue without accountability — must be rejected outright.
It only becomes dogma when people are forced to adhere to rules that have irrational objectives.
Lawyers often get lost in semantic games at this point. How do we know if it is possible? How do we determine it?
The answer is the same way we solve everything else: through rational thinking. This begins with clear definition. A definition establishes meaning by locating something in relation to at least two other things, subject to criteria about what can and cannot serve as a reference, and defines a border.
Language shifts over time, and with it the meaning of words. Meaning is always functional; the way a function is expressed will differ across contexts. We cannot cling to one “fixed” meaning if circumstances change, because preferences and realities change.
Law inevitably becomes dysfunctional because it seeks what cannot exist: a static meaning across all circumstances not even functionally if the criteria is influence by perceived values. Meaning holds only while circumstances remain within a certain range; once they shift too far, the meaning collapses. At that moment, the rule of law ends — it cannot go further. It must stop acting and wait for the demos to endow it with new meaning.
But if law tries to transform meaning itself, it ceases to be law and becomes politics.
The scope of legal operators is blurry, but never undetermined: each fact we decide is relevant to describing a journey becomes part of a legal construct, and we then decide by principle how many “stops” we count and what is worth remembering. This allows us to confidently infer a range of elements that distinguish one construct from another — e.g. what makes a set of facts “murder” rather than “tax obligation” — provided the contrast is sufficient and the criteria objective. This is how the Romans built law, and how any meaningful legal system has worked since.
Our perception of reality is always constrained by uncertainty: we can never be sure we have remembered or observed everything of relevance. To manage this, we use logical placeholders — categories like murder — into which we map perceived acts and the assumptions we derive from them. That mapping is then subjected to a process that checks whether it was done according to agreed methods.
Suppose someone was seen entering a building, and a murder happened five hours later. If this was a large apartment block in New York with a hundred people going in and out every hour, our inference should be very different than if it was a single house in a remote area of Arizona. Does that mean you are less guilty of murder in New York than in Arizona? Of course not. It means the meaning of the observation differs with context. In both cases, the observation alone proves nothing about what happened five hours later.
Uncertainty arises from assumptions and from the sequence in which new information arrives. The law addresses this through a coherence test: it sets a bar. If you can “cha-cha” under the bar, you go free. The law thus constructs an imaginary dance — a compression of noisy data into a structured form that makes facts comparable. Computer science calls this normalisation. Lawyers call it adjunction of law. Computer science has written it down as machine code; lawyers, for reasons of tradition, have not.
This is how uncertainty becomes manageable — not predictable, but governable. This is the essence of the rule of law. Without it, governance collapses into superstition and coercion.
A basic beginner’s guide for lawyering
It is not difficult to understand what the duty is when a lawyer does legal work in a capacity that triggers an obligation to act based on rational arguments, judging in line with the rules and their conscience. If your conscience tells you that applying the rules produces an outcome undermining the stated purpose of those rules, then you must decide whether breaking the rules is required to achieve what the rules are meant to secure.
In a system that can reasonably be described as subject to the rule of law, such situations may arise — because human rights infringements can only be identified by actual humans who are suffering. However, this should not occur as a predictable pattern if the rule of law is truly implemented.
The need to deviate in interpretation cannot be determined without a reasoned position about the constitutional framework in which it arises. If a system operates under the rule of law, it is illogical to consider a perceived unfair outcome in a specific case as justification to disapply the law. The control a judge needs to make is whether the system is a system of law. If not, they cannot enforce it. If it is lawful, then they must enforce it as written — and what is written must mean what was meant. A judge has the responsibility to ensure that their interpretation stays close to that meaning.
The rule of law then has a control mechanism: a person may challenge the system in court, but that duty to test and, if necessary, reshape the law can only be discharged by the highest court of the country. Not every lawyer may assume that responsibility. If they do, they break the coherence of the system and thereby infringe human rights.
Writing encodes meaning. The meaning isn’t in the symbols; it’s in the shared semantic anchors we agree on.
Delta must be small. When I read your words, my reconstruction of meaning should not drift far from what you encoded. Otherwise, communication collapses.
Anchors matter. If I don’t speak Chinese, I can speculate all day, but without anchors (grammar, lexicon, usage), it’s just noise. The same applies to law: without common semantic anchors, interpretation becomes free-floating speculation.
Law requires stable anchors. That’s why treaties must be read according to the ordinary meaning of their terms (Vienna Art. 31). Without that, “interpretation” becomes arbitrary.
Change is possible — but needs to be codified. If we want “yes” to mean “no” sometimes, we must amend the treaty and be specific. That is codification. What we cannot do is say: “They could have meant no after they made the law, so we’ll treat yes as no.” That destroys meaning and the rule of law entirely.
Instead of reasoning from first principles (anchors, meaning, codification), lawyers slip into a style of writing that looks “legal” but it is not. Everyone performs the ritual — citing Vienna, gesturing at “interpretation,” sprinkling “experts say…” — but because nobody has written down the rules of the dance, each one thinks they’re moving in rhythm while in fact they’re stepping on each other’s toes.
It becomes a ballroom of confusion: one’s doing a waltz, another a cha-cha-cha, a third a tango — and because nobody listens carefully to the faint music (‘rule of law’), they believe they’re dancing together.
What you actually get is a room full of collisions, bruised feet, and smiling judges insisting the choreography is flawless. Ouch.
In the meantime…
Me: So what do you say?
Council: Bof, je ne danse pas. 🥱 (“Meh, I don’t dance.”)
Me: I see. 🙄🙄
Coming back to the Assembly recommendations, the brain–computer interface bit piqued my interest because they call on the Committee of Ministers
“to ensure that consideration is given to the possibility of protecting ‘neurorights’ through an additional protocol to the Convention for the Protection of Human Rights and Fundamental Freedoms […] to take into account the potentially unique and unprecedented impact on human rights of the use of artificial intelligence in connection with brain–computer interface (BCI) systems when assessing the feasibility of a legal framework for artificial intelligence.”
It has all the buzzwords and contrived causality: BCIs could shift meaning so dramatically that we’d need to change the definition of human rights. But they have no meaningful technical grounding — not for how AI works, not for how the brain works, and not even for what a brain–computer interface properly is. Any resolution they draft therefore fails basic reasonableness. They could just as well issue recommendations plotting their preferred storyline for the next five Star Wars movies and declare which narratives best align with human rights. But protecting human rights requires reasoned argumentation — not theatrical speculation.
Memorandum paragraph 47:
the integrity of democracy and its processes is based on two important assumptions referred to in Article 7, namely that individuals have agency (capacity to form an opinion and act on it) as well as influence (capacity to affect decisions made on their behalf) Artificial intelligence technologies can strengthen these abilities but, conversely, can also threaten or undermine them. It is for this reason that paragraph 2 of the provision refers to the need to adopt or maintain measures that seek to protect “the ability [of individuals] to freely form opinions”.
They then say it can include
“measures to address the spreading of misinformation and disinformation.”
Suppressing information for the reason they could be misleading is always incompatible with the idea of free decision making. A rule of law system cannot outlaw by a criteria that says an object carries in itself the property of ‘untrue’ or ‘misleading’ since they are relational term: misleading with respect to what? A rule of law system instead defines the process by which a given set of information has to be verified to accept them into a legal situation covered by law (as usual including purpose based on principle and justification why this should be valid). This can include providence, completeness i.e. information are only accepted together with other data not as a single data artefact, the type of person is about etc and much more.
If filters, removals and opaque classifiers start to be the source of our “facts,” people can no longer infer, test, compare or normalise. You end up with a curated hallucination instead of a public record — and once that happens, reason dies.
Content filter imposed by Facebook, Google, Microsoft and everybody else are fundamentally incompatible with the rule of law.
You don’t know what you don’t see.
A content filter removes from view and you don’t know what it keeps from you. Reasoning requires evidence. Rational thinking is impossible if the facts we know of a curated illusion. Removing evidence removes the possibility of testing, rebuttal, and correction — the three pillars of public reasoning.
You don’t know that a ruling was made, or why, or about what — and an algorithm is not neutral: it’s a person deciding what the algo does. Content filtering is a way for the law to subjugate itself to the totalitarian control of whoever controls the machine; and as a result they control everybody else. But the law calls it “safety” or “avoiding misinformation.”
The rule of law has been outmanoeuvred by marketing.
Branding and putting a different sticker on something is enough for the legal system to become a coercive regime that contradicts its purpose, unable to identify when it is faulty, broken and mutates into the disease it was meant to contain. That is what follows when words and laws have no meaning and when yes means no.
But what is the law? It is carried by people. If all those people have lost the ability to act mit Vernunft und mit gewissenhafter Entscheidung — with reason and subject to their conscious - which is the case if they cannot see how wrong this panic about deepfake or online misinformation is — then the law becomes despotic and a threat to the people. This process is already underway.
“The Council of Europe: guardian of Human Rights, Democracy and ... The Council of Europe is the continent's leading human rights organisation. “
They actually write this. The continent's leading human rights organisation?
Can we shop around, see if Australia maybe got something better?
Don’t Anthropomorphise AI
Whoever wrote the following, assuming they understood what it means, gave me some hope that not everything is already lost.
Memorandum paragraph 53: This provision emphasises the importance of human dignity and individual autonomy […] “Activities within the lifecycle of artificial intelligence systems should not lead to the dehumanization of individuals, undermine their agency or reduce them to mere data points, or anthropomorphise artificial intelligence systems in a way which interferes with human dignity.
The problem is that they have a poorly articulated idea of what is required for human autonomy and what follows from this.
Article 7 – Human dignity and individual autonomy
Each Party shall adopt or maintain measures to respect human dignity and individual autonomy […].
Does it mean only adopt or does it encourage new laws?
What criteria apply when it respects dignity and decides how.
On the crucial aspects, their law is mute.
The memorandum mentions that parties to the Treaty should not anthropomorphise AI. I agree, but that renders the EU AI Act incompatible with the Treaty obligations.
And they can thank the OECD for this because the EU incorporated the same irrational definition:
Article 3 — Definitions
‘AI system’ means a machine-based system that is designed to operate with varying levels of autonomy and that may exhibit adaptiveness after deployment, and that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments;
The OECD definition doesn’t just misdescribe AI. It defines something that has never existed in any human technology. Inference only makes sense relative to perception and context, not relative to an already-declared objective.
Perception: I feel hunger.
Objective: Reduce hunger (no longer want to be hungry).
Inference: Given the perception and the objective, I infer which action (pizza, salad, fasting, water) best satisfies the objective.
If the objective is pre-set, then the system doesn’t infer anything — it just maps inputs to outputs.
Their definition is logically dysfunctional. It conflates perception, objective, and output into one messy loop, so it describes no real system. It’s so vague it could apply to anything and nothing. It is not of sufficient quality needed in science. And neither in governance. And it is incompatible with the rule of law because what they define doesn’t exist and therefore anything they want is considered within the scope of the definition.
The act of homicide includes any act of non-homicide; and we will punish homicide. So be careful when you act. That is the EU AI Act.
And it comes from the OECD whose definition talks about implicit objectives, thereby anthropomorphising AI when in fact LLMs have no objectives.
OECD is not a party (and cannot be).
Why: Under Völkerrecht, only States (and certain IOs with capacity) can be parties to a Council of Europe convention. The OECD is neither a State nor a CoE-party IO.
Result: The OECD cannot supply binding treaty content; it has no standing to create obligations inside a CoE treaty.
Observers don’t negotiate treaties.
Why: “Observer” status allows attendance and remarks, not authorship or consent. Treaty text is made by subjects of international law with capacity to be bound.
Result: Anything coming from an observer is, at best, non-binding input. It cannot substitute for negotiated party language.
The OECD AI Recommendation is soft law, not technical standard-setting.
Why: It’s a non-binding policy recommendation (no treaty force), and the OECD is not a scientific SDO (like ISO/IEC) that issues testable, technical norms.
Result: Importing it gives political prose the role of a scientific/legal definition without the legitimacy or rigor of either.
The OECD definition is technically incoherent.
Why: It collapses perception/objectives/inference/output into a catch-all; “machine-based,” “inference,” “adaptiveness,” and “influencing environments” are undefined and overbroad.
Result: A definition that covers everything and nothing makes proportional regulation and rights protection impossible (void-for-vagueness risk).
The treaty lifts that definition essentially verbatim (Article 2).
Why: The CoE text mirrors the OECD wording rather than a bespoke, negotiated legal definition.
Result: Parties appear not to have negotiated scope; they delegated it—informally—to a non-party forum.
This undermines sovereign consent under Völkerrecht.
Why: Treaty obligations rest on the agreement of the parties; meaning must come from what the parties adopted and accepted (Vienna interpretation logic). Here, meaning is imported from an external, unaccountable source.
Resut: The core scope of obligations is not anchored in party consent; that weakens legitimacy and interpretability.
Accountability gap: the OECD cannot be held to account for the words it supplied.
Why: As a non-party, the OECD is outside CoE compliance, court review, or inter-party dispute mechanisms.
Result: States (with consequences to their subjects) are bound by language an outsider wrote, but that outsider bears no responsibility—classic democratic deficit.
Practical consequences
Legal uncertainty: Authorities can label almost any software “AI.”
Arbitrary enforcement: With no hard boundary, discretion—not law—decides.
Signature games: Parties sign immediately or never; it scarcely matters if obligations are indeterminate.
Policy capture: External policy clubs shape binding scope without accountability.
Purpose defeat: A treaty meant to secure human rights/rule of law starts by erasing definitional clarity—undercutting its own aim.
The Council served up Ceremonial Boohai with Legal Flavour.
A treaty that claims to protect human rights, democracy, and the rule of law is built on a definition that voids law itself — no boundary, no test, no anchor.
The parties outsourced consent. They didn’t negotiate meaning; they imported it wholesale from an unaccountable policy club (OECD). That is the opposite of Völkerrecht, which rests on sovereign will.
The result: no substance other than good intentions. The entire structure of Treaty No. 225 is legally incoherent and politically fraudulent.
And the legal experts still claim to uphold the rule of law?
Yeah but no but yeah
Treaty Article 3 – Scope
4 Matters relating to national defence do not fall within the scope of this Convention.
Memorandum paragraph 36:
For the exemption of “matters relating to national defence” from the scope of the Framework Convention, the Drafters decided to use language taken from Article 1, d, of the Statute of the Council of Europe (ETS No. 1) which states that “[m]atters relating to national defence do not fall within the scope of the Council of Europe”. This exemption does not imply that activities within the lifecycle of artificial intelligence systems relating to national defence are not covered by international law.
Me: No, Council 😤 — it means exactly what it says as far as your treatymaking competence is concerned. When it is out, it is out, not in. If you cannot comprehend the difference between yes and no, you should not be drafting or administering matters of law. 🤬🤬 You cannot overrule your own foundational limits by yourself and call that ‘rule of law’ protection. I mean, you can obviously but not legally…
Council: 😴
The Council of Europe’s rationale is absurd. It presupposes that an individual soldier driving a car transforms his professional work, which belongs to “national defence,” into a different domain — as if it would now fall under the "responsible car driving treaty," regardless of the fact that the soldier is not driving to the mall in Strasbourg but from one military station to another. That is not law; it is undemocratic overreach, because it has the consequence that no word, no sentence, no law, no treaty — nothing we can communicate between human beings — has a defined meaning. Instead, it becomes possible that military means military and not military. A human is a human and is not a human. People have rights, and also no rights, all at the same time. That is despotism, and it is precisely what we try to prevent through institutions — not what we should be exporting globally as a totalitarian governance toolkit.
Either the Statute’s defence carve-out excludes anything for the purpose of defence or it does not. You cannot have it both ways — claiming defence is outside the Council’s remit while simultaneously asserting the Council may regulate the same conduct whenever a defence actor relies on anything that could be used in a non-military way, which is essentially everything. This rhetorical arrogance renders Treaty 225 legally incoherent at best, especially since its stated purpose is to promote Human Rights, Democracy and the Rule of Law. In fact, it achieves the opposite and is politically a serious issue that cannot simply be ignored.
And to the German Verfassungsorgane, who have indirectly facilitated the EU’s signature to this Treaty by being members and sponsors of the EU, this cannot be allowed to stand without failing their duty under the German Basic Law (Grundgesetz), which binds Germany to Rechtsstaatlichkeit (rule of law).
They cannot be part of a structure that lacks adherence to the principles of Rechtsstaatlichkeit (rule of law) that pretends to strengthen what must be the law in Germany in any case while corrupting the very meaning of it.