Artificialintelligence

AI Tools

Overview of AI tools

AI tools

Guide to tools for working with artificial intelligence at BUT. The site helps you choose the right tool based on your purpose, type of data, and required level of security. It includes tools supported at BUT, e-INFRA CZ services available to the academic community, GitHub Copilot, selected commercial tools and open-source solutions suitable for the academic environment.

Generative AI is understood at BUT as an assistive and consultative tool. The user is responsible for the lawful use of input data, verification of outputs, compliance with copyright, protection of personal data, protection of trade secrets and transparent disclosure of the use of AI where the rules require it.

Quick tool selection for the job

Recommended direction by type of work
I need Recommended tool or direction Note on use
Chat with a language model, prepare text, summarize materials, tune wording or analyze common documents GPT Chat / Open WebUI e-INFRA CZ or a tool listed internally Generative AI tools at BUT Suitable for general academic and research work. Before uploading a document, always check that it does not contain personal, sensitive, contractually protected or non-public information.
Connect a language model to a script, application, workflow or research pipeline OpenAI-compatible API e-INFRA CZ Endpoint https://llm.ai.e-infra.cz/v1/ je kompatibilní s běžnými klienty pro OpenAI API. API klíč se vytváří v Open WebUI.
Work with data in Python, R, a notebook or learning environment JupyterHub / JupyterLab / RStudio e-INFRA CZ Suitable for teaching, prototyping, data analysis and reproducible computing workflow.
Program with an AI assistant GitHub Copilot for students and teachers, possibly AI Coding Assistants via e-INFRA CZ Before using with non-public source code, check the project rules, licensing regime, working with secrets and repository settings.
Quickly create a website design, prototype or demo page DeepSite e-INFRA CZ, or other generators of websites and presentations Always review the output for security, accessibility, copyright and compatibility with the editorial system.
Create an AI agent or automated workflow n8n AI Agents in e-INFRA CZ Agents can call tools and change data. Always limit permissions, use a test environment, and leave people-impacting decisions to humans.
Run your own chatbot, RAG, local LLM or internal knowledge base Open-source tools, such as LibreChat, RagFlow, LM Studio, AnythingLLM or Open WebUI On-premises or faculty deployments can reduce data risks, but create responsibilities for operations, updates, logs, access, models, and security.
Process personal, sensitive or contractually protected data SensitiveCloud or a pre-assessed safe environment Do not use public chatbots or regular commercial services without a contractual, legal and security assessment.

Basic safety rule

Only enter data into the AI tool that you are authorized to process and that corresponds to the security mode of the given tool. If you are unsure whether a document contains personal data, non-public data, third-party know-how, restricted source code, contractual documents, or grant information, do not use the tool without first evaluating it.

  • Always objectively verify AI outputs. AI may produce inaccurate, incomplete or contrived answers.
  • Indicate the use of AI where it is required by the assignment, publication rules, project rules, license conditions or BUT internal rules.
  • Do not enter BUT's non-public information, personal data, sensitive data, contracts, secrets, non-public source code or third-party data into public and consumer AI services without verified contractual protection.
  • For tools that create images, video, voice, or impersonation of a person, review authorship, personality, and ethical risks and flag synthetic content as appropriate.
  • Use minimum permissions and human approval of critical actions for agents, tools with access to email, calendar, files, repositories, or APIs.

Tools supported at BUT

There is an internal list on the BUT portal Generative AI tools at BUT. These are tools that can be used in the BUT environment and for which, according to the AI tools page, licensing and payment for the tool are solved.

Basic links to BUT tools and catalogs
Source What it contains When to use
AI tools - BUT BUT's public guide to supported, commercial, open-source and e-INFRA tools. As a first reference point for tool selection.
Generative AI tools at BUT Internal list of supported tools with licensing and payment resolved. When choosing a tool for study, teaching, research, administration or operation.
Commercial AI tools Selection of commercial tools and marketplaces that BUT users have experience with. For inspiration and guidance; it is not an automatic approval to work with non-public data.
Open-source AI tools A selection of open-source tools suitable or interesting for the academic environment. For local, faculty or project deployment after technical and safety assessment.
GitHub Copilot BUT information on using GitHub Copilot when working in GitHub and on availability for teachers and students via GitHub Education. For programming, scripting, code learning, and reproducibility support if project rules allow.

AI services of e-INFRA CZ and CERIT-SC

As members of a research organization, students and employees of BUT can apply to the MetaCentre and use the services of the national e-infrastructure e-INFRA CZ. The CERIT-SC documentation describes AI as a Service as an on-premise AI platform for researchers, accessible via Open WebUI and OpenAI-compatible APIs.

AI services of e-INFRA CZ relevant for BUT
Service Usage Access and security notice
GPT Chat / Open WebUI Chat, work with text, code, documents, image functions, RAG over embedded documents and model selection. Requires a valid MetaCentra account. The service runs in the e-INFRA CZ environment; However, the WebUI is not automatically designated for protected, classified, or personally identifiable data without further consideration.
OpenAI-compatible API Connecting LLM to scripts, applications, pipelines, backends, LangChain frameworks, LlamaIndex or custom services. The basic endpoint is https://llm.ai.e-infra.cz/v1/. API klíč je nutné chránit jako tajemství.
JupyterHub / JupyterLab / RStudio Notebooks, learning programming, data analysis, Python, R, machine learning and reproducible computing. Suitable for academic and research workflow. For datasets with personal or protected data, choose the appropriate security mode.
AI Coding Assistants Integration Integration of e-INFRA models into development environments and coding assistants. Suitable for developers who want to use locally operated models instead of external commercial providers.
DeepSite Generating web pages and applications from text input. Requires a MetaCentrum account. Outputs must be checked before publication, especially from the point of view of security, accessibility and rights.
n8n AI Agents Visual creation of agents and automated workflows with connection to language models and other tools. Use least privilege, a test environment, auditable logs, and human approval of actions impacting people, finances, data, or systems.
CERIT-SC Kubernetes Container Platform Operation of containers, web services, computational workflows, custom models, databases and application prototypes. Access is available to the Czech academic community via the e-INFRA CZ / MetaCentrum identity. A project quota is requested for intensive resources.
SensitiveCloud Secure computing and storage environment for sensitive research data. Suitable for projects that require a higher security mode. Resources are requested through the principal investigator of the project.

Programming, development and working with code

To work with code, you can use GitHub Copilot, coding assistants connected to e-INFRA models, or local or open-source solutions. For non-public code, it is necessary to verify whether the terms of the project allow it to be inserted into the tool, whether the tool uses the code for training, how it stores the context and how it protects the access data.

Commercial AI tools

The BUT public page contains an overview of commercial tools and marketplaces. This overview is intended as inspiration, not as automatic approval to work with internal, personal or proprietary data. Before using a commercial tool, check the contractual terms, license regime, protection of embedded data, use of data for training, place of processing, subcontractors, data retention and the possibility of data deletion.

Categories of commercial tools listed on the BUT portal
Category Examples of tools or resources Safety note
AI tools marketplace topai.tools, Kifinity, Future Tools, AllThingsAI, AI Tool Hunt, There's An AI For That, awesome-ai-tools, Papers with Code, AIXploria The marketplace is used to find the tool; each tool must be assessed separately.
GPT and general chat tools ChatGPT, Claude, Gemini, Microsoft Copilot, Mistral, Groq, Grok, Cohere Coral, Perplexities, NotebookLM For public or low-risk data; for internal and personal data only in licensed or contractually assessed mode.
Research and scientific resources Perplexities, NotebookLM, Elicit, Scythe, SciSpace, Semantic Scholar, Scinapse Always verify citations and claims with primary sources. Do not post non-public manuscripts, grants, or review materials without permission.
Image generation Midjourney, DALL·E Check copyright, rights of persons, identifiability of persons, license terms and labeling of synthetic content.
Video generation Runway, Veo, Sora, NotebookLM.video Address consent, labeling and reputational risks for deepfakes, realistic depictions of people or public communications.
Voice generation Eleven Labs Do not impersonate a specific person's voice without authorization and consent; there are personal, security and reputational risks.
Documents, texts and presentations Gamma, Scribe, Writesonic, QuillBot, Anyword, Slidebean, Beautiful.ai Only enter data that can be processed by the given service. Check the output for factual accuracy, rights and plagiarism.
Code and development Qodo / Codium, CodeSignal, Replit Risky with non-public code, secrets, access tokens, contract development and security sensitive projects.
LLM and data processing Datrics, Sisense, DataRobot It is often about processing datasets; requires consideration of legal title, contract, processing relationship, security and, where appropriate, DPIA.

Open-source and locally run tools

An open-source tool does not automatically mean a secure tool. The advantage can be local operation, control over data and the possibility of customization. At the same time, it is necessary to ensure update management, secure settings, accesses, logging, backups, license control, selected model and operational responsibility.

Open-source tools listed on the BUT portal
Tool Typical use What to check before deployment
LibreChat Chat interface for both commercial and local language models. Authentication, user management, logs, connected models, data location, backups.
Perplexica Research Search and search interface inspired by Perplexity. Resource quality, search integration, sending queries to external services, local model.
RagFlow RAG, Knowledge Bases, Agents and Chatbots. Work with documents, access rights, indexes, deletion of data, separation of projects.
LM Studio Working with local LLM models on a work computer. Hardware requirements, model origin, model license, updates, data handling.
AnythingLLM Local or server interface for LLM and knowledge base. Document indexing, access, logging, connected embedding models and data deletion.
Transformer Lab Experiments with local and commercial models including fine-tuning. Licensing regime of models and data, security of training data, reproducibility.
Open WebUI A chat interface for both commercial and local models, suitable for custom deployment. Server traffic, authentication, logs, retention policy, model access and deployment security.

Recommended mode by data type

Indicative choice of tool according to data sensitivity
Data type Examples Recommended mode Inappropriate mode
Public data Published web texts, public articles, open datasets, public annotations. Common AI tools after verifying the conditions and output quality. No confidentiality restrictions, but copyright and citation still apply.
Internal but not sensitive data Internal methodologies, working concepts, unpublished presentations without personal or contractually protected data. Tools supported at BUT or e-INFRA CZ; commercial services only after verification of the contractual terms. Public AI tools without contractual safeguards, unless the data is intended for publication.
Personal data Data of students, employees, applicants, research respondents, contacts, study or work data. Only after verification of the legal title, purpose, processing relationship, security and possibly consultation with the personal data protection officer. Consumer chatbots and tools without clear contractual protections and processing controls.
Sensitive, special or contractually protected data Health data, biometric data, research data under NDA, trade secrets, partner's non-public know-how, security information. An individually assessed environment, such as SensitiveCloud or another approved secure mode. Common web AI tools, common cloud services without contract and security assessment.
Confidentiality, Access Data and Security Information Passwords, API keys, private keys, tokens, configurations, vulnerabilities, incident logs. Do not invest in AI tools; work in a security-approved mode and according to CIS/CSIRT rules. Any tool that is not expressly authorized to process this type of data.

How to assess or implement a new tool

  1. Describe the purpose. What will the tool be for, who will use it and what decision or output will it support.
  2. Specify the data type. Differentiate between public, internal, personal, sensitive, contractually protected and security data.
  3. Verify provider and terms. Check license, price, contract, data processing, input training, retention, export, deletion, subcontractors and place of processing.
  4. Perform a rapid screening under the EU AI Act. Pay particular attention to education, testing, admissions, HR, employee evaluation, biometrics, emotion recognition, public communications and synthetic content generation.
  5. Involve the responsible people. Depending on the nature of the use, consult the AI contact point, CIS/CSIRT, data protection officer, legal support, project sponsor or data owner.
  6. Test on low-risk data. First, use public or test data to verify accuracy, stability, bias, citations, safety, and cost.
  7. Write down the decision. Include the approved purpose, limitations, permitted data types, responsible person, review date and link to the terms of the tool.

Use the follow-up page for a systematic review of the tools Risk and security assessment of AI tools.

Verified links

Responsible person: doc. Ing. Petr Koňas, Ph.D.

Responsibility: doc. Ing. Petr Koňas, Ph.D.