AI deployment knowledge base: regulations, templates and comparisons
Materials with a legal basis for every date: what the Polish AI Systems Act says, how to write an AI use policy, how to get shadow AI under control and which models to choose
Looking for a definition or a short answer? AI glossary · Questions and answers
Polish AI Systems Act: what companies and public bodies must prove from 28.10.2026
Timeline, KRiBSI oversight, duties and fines - each date with its legal basis
Timeline 2025-2028: when the next requirements take effect
Regulations and AI policy
6 materialsPolish AI Systems Act
What companies and public bodies must prove from 28 October 2026: timeline, KRiBSI, duties and fines
→AI use policy for companies - template
11 sections, a list of data that must not be entered, a tool register; downloadable template
→Shadow AI
What it is, how widespread, the risks to data and liability, and 5 steps without a ban
→Council session transcription
From the session recording to draft minutes and BIP captions - in the office’s infrastructure
→KSC amendment (NIS2) and AI
Who is covered, the deadlines and what to show about AI tools - with a status check tool
→KNF guidelines and DORA for AI
What banks, insurers and fund management companies must show about AI tools - with an obligations calendar
→AI models and environments
2 materialsTools - quick check
8 mini-tools in the materialsInstant result, plus a Glossary term, a calendar date and the document your answer produces
What you must be able to prove by 28.10.2026
Type of organisation and AI use → duties and evidence
System register · literacy · no prohibited practices · rules
Is your AI system high-risk
Annex III area, impact on a decision about a person, Art. 6 exception → risk class and duties
Art. 26 duties from 2.12.2027 - oversight, logs, impact assessment
Check your policy
Which of the 11 template sections are already in place - and what to add first
Most urgent: Labelling AI content - Art. 50 from 2.08.2026
How much shadow AI is in your organisation
Headcount, company tool, policy → an estimate of people outside control
First step: inventory and a safe tool
Can you spot an AI hallucination
Two answers to the same question - which is made up? Your score and why a source citation matters
The made-up answer had no source citation - as in TWIN:DESK (RAG: an answer with a citation from documents)
Check the captions deadline
The session date and the office’s choice give the captions deadline and the latest deadline
25.10 captions deadline
Does the KSC Act apply to your organisation
Type of organisation, sector and size give the status, the deadlines and the first step
Important · ENTITY
Calendar of AI obligations for your institution
The type of institution and how you use AI give a calendar: what already applies and what is coming
NEXT DEADLINE 2.12.2027
Educational games
5 games + Daily PunA few minutes of play, with explanations of AI terms, AI Glossary entries and ideas for training sessions
How much data can you keep from leaking to AI
Public office, bank, hospital, energy → protect data, block attacks, let ordinary work through
PROXY:AI does the same automatically: in under 50 ms, in 12 languages
What 100 people said when asked about AI
Miners, chefs, doctors → 4 answers in every pun, the game understands meaning
Junior: pick 3 of 4 answers. Expert: answers in your own words
Who are you talking to: a human or a machine?
A bank, a “grandson”, a helpline → real lines from the press and AI answers, then a safe decision
Street poll: the machine among three answers. Phone: five calls, two decisions in each
When AI just agrees
Report, application, bid → three prompts, three answers, one echo
A prompt that assumes the result gets the result. We ask for arguments for and against
When does an AI model become an agent?
Letter, offer, lease → the same horse without a harness and in a harness of 6 parts
Same model, different result: accuracy, time, oats and slips side by side
