AI compliance, proven. For humans and their agents.
Every law that touches your AI. Cited.
Perissos gets your AI and your agents ready for the laws that bind them, proves it with sealed evidence any buyer can verify without asking us, and reads any vendor's AI from what they publish.
Check your AIFree, no signup, about three minutes. Works on any company, even one that has never heard of us.
Where these dates come from
- newly added banned practices: Article 113, third paragraph, point (a), as replaced by Regulation (EU) 2026/1744, Article 1, point (40)(a). New prohibitions with their own later date, carved out of the Chapters I and II phase.
- marking of AI-made content by systems already on the market: Article 111(4), added by Regulation (EU) 2026/1744, Article 1, point (39)(b). This is relief for a legacy cohort, not a postponement of Article 50: it reaches providers only (not deployers), systems generating synthetic content only, systems placed on the market before the general application date only, and Article 50(2) only. Article 50(1), (3) and (4) are untouched, and systems placed on the market from the general application date onward comply on that date.
560 cited provisions. 7 EU sources monitored for amending instruments.
Our own readiness, replayed from the sealed snapshot.
Generated 28 August 2026. 1 of 1 EU AI Act items shown as published.
No sign-up. Your result is an automated draft, not legal advice.
every evidence file sealed with SHA-256. exports signed with Ed25519, verifiable offline.
As of 30 August 2026. EU AI Act: 8 of 15 obligation groups in force, 7 still to apply. Next on 2 December 2026, in 94 days: newly added banned practices and marking of AI-made content by systems already on the market. 560 cited provisions, 22 official sources, 9 monitored for change.
Where these dates come from
- first banned practices and general rules: In force since 2 February 2025 (Regulation (EU) 2026/1744 (Digital Omnibus on AI); Regulation (EU) 2026/1744, Article 1; ELI http://data.europa.eu/eli/reg/2026/1744/oj; CELEX 32026R1744)
- general-purpose AI model duties, governance and penalties: In force since 2 August 2025 (Regulation (EU) 2024/1689 (Artificial Intelligence Act); Article 113; ELI http://data.europa.eu/eli/reg/2024/1689/oj; CELEX 32024R1689)
- changes to other EU laws: In force since 27 July 2026 (Regulation (EU) 2026/1744 (Digital Omnibus on AI); Regulation (EU) 2026/1744, Article 1; ELI http://data.europa.eu/eli/reg/2026/1744/oj; CELEX 32026R1744)
- transparency duties: In force since 2 August 2026 (Regulation (EU) 2024/1689 (Artificial Intelligence Act); Article 113; ELI http://data.europa.eu/eli/reg/2024/1689/oj; CELEX 32024R1689)
- fines for general-purpose AI model providers: In force since 2 August 2026 (Regulation (EU) 2024/1689 (Artificial Intelligence Act); Article 113; ELI http://data.europa.eu/eli/reg/2024/1689/oj; CELEX 32024R1689)
- conformity assessment, CE marking and registration of high-risk systems: In force since 2 August 2026 (Regulation (EU) 2024/1689 (Artificial Intelligence Act); Article 113; ELI http://data.europa.eu/eli/reg/2024/1689/oj; CELEX 32024R1689)
- older AI components of large-scale EU IT systems: Applies from 31 December 2030 (Regulation (EU) 2024/1689 (Artificial Intelligence Act); Article 113; ELI http://data.europa.eu/eli/reg/2024/1689/oj; CELEX 32024R1689)
- older high-risk systems used by public authorities: Applies from 2 August 2030 (Regulation (EU) 2026/1744 (Digital Omnibus on AI); Regulation (EU) 2026/1744, Article 1; ELI http://data.europa.eu/eli/reg/2026/1744/oj; CELEX 32026R1744)
- older general-purpose AI models: Applies from 2 August 2027 (Regulation (EU) 2024/1689 (Artificial Intelligence Act); Article 113; ELI http://data.europa.eu/eli/reg/2024/1689/oj; CELEX 32024R1689)
- registration in the EU database: In force since 2 August 2026 (Regulation (EU) 2024/1689 (Artificial Intelligence Act); Article 113; ELI http://data.europa.eu/eli/reg/2024/1689/oj; CELEX 32024R1689)
- post-market monitoring, incident reporting and the right to an explanation: In force since 2 August 2026 (Regulation (EU) 2024/1689 (Artificial Intelligence Act); Article 113; ELI http://data.europa.eu/eli/reg/2024/1689/oj; CELEX 32024R1689)
- newly added banned practices: Applies from 2 December 2026 (Regulation (EU) 2026/1744 (Digital Omnibus on AI); Regulation (EU) 2026/1744, Article 1; ELI http://data.europa.eu/eli/reg/2026/1744/oj; CELEX 32026R1744)
- marking of AI-made content by systems already on the market: Applies from 2 December 2026 (Regulation (EU) 2026/1744 (Digital Omnibus on AI); Regulation (EU) 2026/1744, Article 1; ELI http://data.europa.eu/eli/reg/2026/1744/oj; CELEX 32026R1744)
- high-risk duties for systems in the Annex III use areas: Applies from 2 December 2027 · moved from 2 August 2026 by Regulation (EU) 2026/1744 (Regulation (EU) 2026/1744 (Digital Omnibus on AI); Regulation (EU) 2026/1744, Article 1; ELI http://data.europa.eu/eli/reg/2026/1744/oj; CELEX 32026R1744)
- high-risk duties for AI inside Annex I regulated products: Applies from 2 August 2028 · moved from 2 August 2027 by Regulation (EU) 2026/1744 (Regulation (EU) 2026/1744 (Digital Omnibus on AI); Regulation (EU) 2026/1744, Article 1; ELI http://data.europa.eu/eli/reg/2026/1744/oj; CELEX 32026R1744)
Watch the whole journey.
Two sides, one engine. Sellers walk from a plain-English description to audit-ready proof. Buyers walk from a web address to verified evidence. Click any step.
1. Say what your AI does
Plain English in. No account, no call. The narration reports each real step as it runs.
Extracting the facts your words state
Nothing here is an opinion.
Every result traces to a pinned law, a sealed file or a signed record. That is the whole machine.
Accredited bodies do the certifying. This engine makes sure that, when they arrive, nothing is missing.
Agents, kept in line. Before, during and after production.
An agent is not a chatbot. It acts. So the rules have to reach it at every stage of its life, and every decision has to leave a record an auditor can hold.
Today: agents register on the registry with an owner, declared tools and oversight, lifecycle recorded, evidence attachable and sealed.
The intent: a decision point in production, where an agent asks before it acts and every answer lands as sealed evidence. That surface is not built yet, and nothing on this page pretends it is.
Read from the outside. Tested from the inside.
Anyone's agents, read from what they publish. Your own agents, tested as hard as an attacker would, with your written say-so.
The outside-in read
- Composition. What it is built from, as published.
- Reach. What its declared tools could touch.
- Integrity. Whether its declarations hold together.
- Behaviour. How it conducts itself at our own door, where we may look.
- Drift. What changed since the last read, with both dated quotes.
Read-only, always. Our crawler honours robots.txt and publishes its own methodology. What we cannot verify renders unverified, never good.
The only outside-in agentic read we know of that is compliance-grade, score-free and two-sided.
The consented red team
The intent: with written authorisation, agents get tested the way an attacker would, and findings land as sealed, dated evidence rows, ready for the Gate and for your buyers. Not yet a feature.
Status anyone can read. Never a score.
The same sealed facts, rendered two ways. Flip it. Nothing changes underneath.
Our own published snapshot, served as sealed. 28 August 2026.
Our risk tier under the EU AI Act, by our own published, sealed snapshot: Limited risk.
0 items are evidence-backed; 0 are self-attested. Unproven claims say so.
No grade, no letter, no number out of 100. The law's own words and what is actually proven.
What the law provides for
These are the maximum fines the Regulation provides for. Member States lay down the rules on penalties and decide each case.
Article 99(1), (7) and (9)
- For the banned practices in Article 5, the law provides for administrative fines of up to EUR 35 000 000 or, for an undertaking, up to 7 % of its total worldwide annual turnover for the preceding financial year, whichever is higher. (Article 99(3))
- For the operator and notified-body obligations Article 99(4) lists, the law provides for administrative fines of up to EUR 15 000 000 or, for an undertaking, up to 3 % of its total worldwide annual turnover for the preceding financial year, whichever is higher. (Article 99(4), as amended: the list now includes the Article 25(2) and (4) obligations, point (da), inserted by Regulation (EU) 2026/1744, Article 1, point (38)(b))
- For supplying incorrect, incomplete or misleading information to notified bodies or national competent authorities, the law provides for administrative fines of up to EUR 7 500 000 or, for an undertaking, up to 1 % of its total worldwide annual turnover for the preceding financial year, whichever is higher. (Article 99(5))
- For small and medium-sized enterprises, including start-ups, each of these fines is capped at the lower of the percentage or the amount (Article 99(6)). For small mid-cap enterprises, the fines in paragraphs 4 and 5 are capped the same way (Article 99(6a), inserted by Regulation (EU) 2026/1744, Article 1, point (38)(c)). (Article 99(6) and Article 99(6a))
Every search makes the next one stronger.
1. A buyer searches
any company, by web address
2. The vendor is invited
claiming their page costs nothing
3. The vendor publishes
signed proof, verifiable offline
4. The next search
shows evidence, not promises
Two-sided by design: the searched become the published.
The whole plan, in six moves.
The autonomous engine
Everything automatable, automated. The few acts the law reserves for a human arrive prepared.
Proof that sells
Readiness becomes signed, checkable proof buyers act on.
The two-sided loop
Buyers search. Vendors publish. The lookup is free.
Legible status
Plain English or full technical. The same sealed facts. Never a score.
AI-native throughout
Agents are users too, under the same rules. Our own assistant runs the product and is registered on our own registry.
The autonomous business
Published prices, self-serve, no call required.
The assistant is live for signed-in workspaces today: it answers with citations and runs the product on your instruction, and nothing changes until you confirm.
Start free. Stay published.
The check
€0
The full readiness check, cited. No signup, no card.
Pro
€249 a month for two months, then €2,490 a year
For one company getting audit ready.
Scale
€499 a month for two months, then €4,990 a year
For heavier volume and a larger vendor portfolio.
Prices published, not negotiated. VAT added at checkout. See pricing →
Real output, not product art
Both screenshots are real product output, run on our own company. Cropped to fit; nothing altered.
Why you can trust the result
Every claim below is checkable.
- Cited to the official texts and reviewed before release.
- 560 cited provisions, quoted, not paraphrased.
- Sources hash-pinned (SHA-256) to the published texts.
- Nine US AI laws monitored for change; EU sources pinned and dated.
- About other companies we publish facts and a verification status, never a score.
- EU-hosted; text sent for AI analysis goes to a non-EU provider, never your evidence files.
Who it's for
Perissos is for companies that build or use AI and sell into the EU or US, where a buyer's review can take 8 to 16 weeks and procurement asks hard, cited questions.
We make you audit-ready and produce the cited proof your buyers ask for. We don't certify or audit you. That's the job of accredited auditors and CPA firms.
See which rules apply to you
Run your free readiness checkIn partnership with Munster Technological University and MTU CyberInnovate.
Perissos was built on the CyberInnovate programme at Munster Technological University.
