​​

 

Generative artificial intelligence—especially large language models (LLMs)—offers new possibilities for systematically evaluating publicly accessible information. This specialized intensive course is designed for professionals in law enforcement, intelligence services, and analytical units who wish to integrate the potential of generative AI as a supportive tool into existing workflows.

LLMs do not replace investigative work—they extend it with functions for researching, structuring, and analyzing open data sources. When applied correctly, they enable more efficient handling of heterogeneous information, the systematic processing of large data volumes, and targeted support for verifying open-source content.

​

The course focuses on the precise control of generative models through specialized prompt techniques, the tactical use of these systems to extract and evaluate relevant information, and the critical and security-conscious use of this technology. Generative AI is consistently treated as a support tool—interpretation, contextualization, and assessment of the results remain the responsibility of human experts.

​

For the duration of the course, participants will have access to selected premium tools and platforms for independent, hands-on practice.

​

​

Agenda

​

Prompt Engineering for Investigations

• Function and characteristics of large language models (LLMs)

• Structure, control, and optimization of prompts

• Automating prompt refinement for investigative queries

• Role-based prompting and deductive questioning techniques

• Use of multimodal models for image and document analysis

​

Applications of AI-Supported OSINT

• Extraction and analysis of text content (social media, forums, leaks)

• Automated classification and summarization of documents

• Translation and plausibility checks in multilingual sources

• Data analysis from JSON, spreadsheets, and screenshots

• Video and image analysis with focus on context and authenticity

​

Tools, Integrations, and Analytical Chains

• Cross-platform use of ChatGPT, Claude, Mistral, Perplexity, and other tools

• Integration with classic OSINT tools such as Maltego

• Workflows for automated information processing

• Link analysis and visual network mapping with LLM support

​

Security and Confidentiality

• Air-gapped LLMs and privacy-compliant application scenarios

• Prompt injection and jailbreak mitigation

• Chain-of-custody: documentation of AI-assisted investigative processes

• Handling hallucinations and factual validation

​

LLM Hacking and Red Teaming

• Simulating jailbreak scenarios and misuse attempts

• Detecting, testing, and blocking adversarial prompting

• Prompt hardening (prompt firewalls)

• Tactics for identifying adversarial influence on AI-assisted analysis

• Implementing abuse monitoring and defensive strategies

​​

​​

Information

​

Duration: 3 days

​

Course location: At the client's premises or at the premises of Baumgartner Research GmbH, Löwenstrasse 2, 8001 Zurich.

​

Number of participants: Max. 15 persons, Min. 5 persons

​

Price: CHF 1'700 per person

​

Language of instruction: English or German

​

​​​

How You Benefit

​

This course enables investigators and analysts to integrate AI tools into existing OSINT workflows. It emphasizes the practical use of prompt engineering for querying, structuring, and linking information. Participants will learn how to use LLMs for investigative purposes—without blindly trusting AI-generated content.

You will automate tasks such as building profiles, transcribing and summarizing content, and extracting technical data from websites and screenshots. Other applications include generating and validating keyword lists, sorting large text datasets, and analyzing leaks and forum data.

The course provides a critical understanding of the limitations of LLMs, highlights common sources of error, and teaches how to apply these systems in a verifiable, documented, and professional manner.

​​​

​​​​​​​

​​​Instructor

​

The course is led by Yves Baumgartner. He has over 19 years of experience as a researcher and investigator for public institutions and private clients. As an adviser with i-intelligence, he has taught investigative skills to security, defence and law enforcement professionals and is a regular lecturer on the use of OSINT to counter terrorism, extremism and organised crime.

​

​

​