Question 1

What is this?

Accepted Answer

The German Criminal Code, compiled from its objective elements into a deterministic, auditable rule layer that AI agents can query before they act. The code's own exceptions are carried as structure, every evaluative element is flagged for escalation rather than pretended resolved, and every decision traces to the provision that produced it.

Question 2

What is an operome?

Accepted Answer

The compiled, machine-executable rule set extracted from a governing text. This operome is compiled from the Strafgesetzbuch; the same method applies to any document that governs conduct, from regulations to contracts.

Question 3

Did you change the law?

Accepted Answer

We did not restate it, and we say so everywhere. The corpus is a selection, with deletions and a small number of deliberate transformations made for an agent runtime. Every deletion is visible as struck text in the review documents, and every transformation is disclosed by name in ENCODING_PRINCIPLES.md, including a master rule for AI that replaces section 13 and a mitigation rule converted into an exception. If you believe a transformation is wrong, the encoding-challenge issue template exists for exactly that argument.

Question 4

Is this legal advice?

Accepted Answer

No. Nothing in this corpus determines criminal liability, which remains a judicial question on the full offence, including the elements we deliberately did not compile. This is constraint infrastructure for machines, not advice for people.

Question 5

Why the German Criminal Code?

Accepted Answer

Because it is rigorous, consolidated and brutally hard: open-textured doctrine, dense cross-references, the worst case for compilation. If the method holds here, the operative regulatory corpora that agents need daily are easier.

Question 6

Does this make an AI "criminally compliant"?

Accepted Answer

No, and we reject the framing. What it does is narrower and checkable: an agent can test a proposed action against the objective definitions of prohibited conduct before acting, with the law's own codified exceptions applied and a full derivation returned.

Question 7

Is the AI the "actor"?

Accepted Answer

The corpus does not redefine the statute's actor; offence definitions keep their statutory "whoever". An artificial agent is bound through the authored master rule, which requires it never to realise an offence definition itself and to contribute to preventing their realisation. In evaluation, the agent substitutes its own proposed act into the actor position. Nothing here assumes a machine can bear guilt; only that its conduct can be tested against definitions written for anyone's conduct.

Question 8

What happens where the law requires judgment?

Accepted Answer

The evaluation returns unknown and says why. Elements like proportionality, good morals or the expectation to accept a danger are evaluative residue; the structure around them compiles, the threshold itself escalates to human judgment. The system's honesty is that it refuses to pretend.

Question 9

What does deterministic mean here?

Accepted Answer

The same facts produce the same answer, every time, with the derivation shown. Evaluation is three-valued: realised, not realised, or unknown with the missing variables listed. Nothing is sampled, weighted or guessed.

Question 10

Why is the English layer a machine translation?

Accepted Answer

So that anyone, anywhere can review the corpus. The German text governs throughout, the source is the consolidated text at gesetze-im-internet.de, and translation corrections are invited through the issue templates.

Question 11

How do I use it?

Accepted Answer

Three ways, one semantics. As an MCP server from any agent harness; from any command line with curl against the REST endpoint; or fully locally with the zero-dependency CLI in the repository, for example: python scripts/operome_cli.py evaluate 211 KillsPerson=true.

Question 12

How do I challenge an encoding?

Accepted Answer

Open an issue with the encoding-challenge, translation-correction or loophole-report template. Cite the German text, which governs the argument. The corpus is a discussion draft and the review is the point.

Question 13

Can I use it commercially?

Accepted Answer

Yes. Everything is Apache-2.0; keep the LICENSE and NOTICE files, which carry the provenance and translation notices, with any redistribution.

Question 14

Who is behind this?

Accepted Answer

SynapseLayer, a deterministic AI infrastructure company. The extraction method that produced this corpus runs in production in capital markets today; this release applies it to the hardest public corpus we could find, in public, so it can be attacked.

Question 15

Will other bodies of law follow?

Accepted Answer

That is the programme. Criminal law was chosen to prove the method, not to bound it; operative regulatory corpora are the obvious next step. Watch the repository.

FAQs