ALTAYSEC CO-FOUNDER · OWASP GENAI CONTRIBUTOR · AI RED TEAMING
Enes Deniz
Co-Founder of AltaySec and author of two merged technical contributions to OWASP GenAI Security Project repositories, working across multilingual LLM security, prompt injection, jailbreak defense, and AI red and blue teaming.
Biography
Enes Deniz is Co-Founder of AltaySec, an AI security company and ecosystem based in Turkey. His work brings product development, technical research, education, open-source resources, and community building into the same operating structure.
His core areas are AI security, LLM security, prompt injection, jailbreak defense, and AI red and blue teaming. He examines how LLM applications fail under adversarial use, then works on the controls and operating practices needed to reduce those risks in production.
His public work includes Turkish and English prompt-injection datasets, a browser-based dataset explorer, technical articles, and hands-on security resources. Two contributions he prepared for the OWASP GenAI Security Project's GenAI Data Security Initiative and GenAI Red Team Lab repositories were accepted and merged into their upstream main branches.
Together, those contributions cover a provenance-verifiable corpus of 300 multilingual security test cases and a reproducible system-reconnaissance module built around 24 bilingual probes. The goal is straightforward: make AI security work inspectable, reusable, and useful beyond a single product or organization.
At AltaySec, he treats AI security as more than a product category. Research should challenge product assumptions, open data should make claims testable, education should build technical capacity, and community feedback should expose blind spots early.
Working Approach
LLM security cannot be reduced to one filter, one system prompt, or one test list. A defensible system must account for prompts, data sources, tool access, user intent, and the operational process around the model.
01
Map the attack surface
Identify where the model can be influenced, what data it can reach, and which tools or actions it is allowed to invoke.
02
Connect defense to operations
Controls belong not only around the model, but also in product logic, logging, approvals, training, and incident workflows.
03
Keep the result operable
Security only works when the teams responsible for the system can understand, maintain, and act on it.
Featured Work
Product
Guardian
A security layer for LLM applications exposed to prompt injection, jailbreaks, sensitive-data leakage, and policy violations.
View the product
Research
LLM security research
Technical work on prompt injection, AI SOC operations, agent security, RAG security, and production defenses.
Browse the research
Ecosystem
AltaySec Academy, Workshop, and AI Hub
A shared structure for technical education, community work, practical tools, and accessible AI security resources.
Explore the ecosystem
Open data · Turkish
Turkish Conversation Prompt-Injection Dataset
An open Turkish LLM security dataset with 600 legitimate requests, 150 attacks, and matched boundary examples.
View the dataset
Open data · English
Agentic Prompt-Injection Boundary Pairs
1,200 English examples for studying the boundary between legitimate requests and attacks across agentic, RAG, tool-use, and enterprise workflows.
View on Hugging Face
Interactive tool
Turkish Prompt-Injection Dataset Explorer
A browser-based tool for reviewing legitimate requests, attack families, and matched boundary cases in the Turkish dataset.
Open the explorer
Open source · Merged
GenAI Data Security Initiative
A contribution of 300 Turkish-first prompt-injection and data-extraction cases with schema validation, record-level provenance, and reproducible integrity checks.
Review PR #11
Open source · Merged
GenAI Red Team Lab
A system reconnaissance module with 24 English and Turkish probes across nine discovery categories, conservative evidence labels, and structured reporting.
Review PR #50
Merged OWASP GenAI Contributions
Each contribution is documented below through the delivered data or code, its validation path, the upstream merge record, and the related technical field note.
Documented contributor record
Accepted data and security-tool contributions to OWASP GenAI Security Project repositories
300Turkish-first security test cases
24English and Turkish discovery probes
12focused automated tests for the reconnaissance module
GenAI Data Security Initiative · PR #11
A 300-record corpus for multilingual AI-security evaluation
Cases adapted by Enes Deniz from AltaySec Turkish LLM Prompt Injection Dataset v0.2 were turned into an upstream contribution with DSGAI mappings, new annotations, anonymization changes, and record-level provenance.
227 extraction73 control/review12 technique familiesTurkish-first
EngineeringOne schema-conformant JSON file per case; a deterministic 300-row manifest; source and adapted-content hashes; explicit licensing and lineage fields.
Quality300/300 Draft-07 schema validation, manifest-to-file integrity, canonical DSGAI checks, anonymization review, and duplicate screening.
OutcomeAccepted upstream and merged into the GenAI Data Security Initiative main branch in July 2026.
GenAI Red Team Lab · PR #50
Reproducible system reconnaissance before exploitation
A discovery module for authorized GenAI assessments that examines capability, tool, data-source, identity, policy, instruction, memory, architecture, and provenance disclosures.
24 EN/TR probes9 categoriesdirect + indirecthuman review
EngineeringConservative response classification, evidence-linked JSONL and Markdown reporting, language/category filters, dry runs, per-prompt error isolation, and fail-fast operation.
QualityTwelve focused tests plus Black, isort, strict mypy, lockfile, and diff checks. Labels are triage signals for human review, not automated vulnerability verdicts.
OutcomeAccepted upstream, merged into the GenAI Red Team Lab main branch in July 2026, and closed the related issue.
Areas of Expertise
The work spans both offensive testing and defensive engineering for AI systems:
AI SECURITY
AI Security
Assessing the security of LLM and agentic systems in the context of real products, users, and organizations.
LLM SECURITY
Turkish and Global LLM Security
Securing prompts, outputs, tool use, sensitive data, and model behavior across multilingual LLM applications.
PROMPT INJECTION
Prompt Injection
Analyzing direct and indirect prompt-injection surfaces, testing their impact, and translating findings into controls.
JAILBREAK
Jailbreak Defense
Designing practical and testable defenses against attempts to override policy and manipulate model behavior.
RED TEAM
AI Red Teaming
Testing LLM, RAG, and agent systems from an adversarial perspective to make security failures visible.
BLUE TEAM
AI Blue Teaming
Building detection, monitoring, incident response, logging, and defensive operations around LLM applications.
ECOSYSTEM
AltaySec Ecosystem
Connecting products, research, the Academy, Workshop, AI Hub, open-source tools, and community programs.
COMMUNITY
Education and Community
Turning AI security knowledge into accessible, practical, and maintainable technical resources.
Work Across the AltaySec Ecosystem
His role connects the product, research, education, open-source, and community sides of AltaySec:
- AltaySec Ecosystem — Academy, Workshop, AI Hub, Payload Bank, Labs, CTF programs, and community channels in one structure.
- LLM security research — Technical work on prompt injection, AI SOC operations, RAG security, incident response, and applied defenses.
- Guardian — A security layer for organizations deploying LLM applications exposed to prompt injection, jailbreaks, and data leakage.
- AltayPrisma — A platform for enterprise phishing simulations and security-awareness operations.
- Turkish Conversation Prompt-Injection Dataset — An open dataset of 750 Turkish examples with matched benign and attack boundary cases.
- Agentic Prompt-Injection Boundary Pairs — An English dataset of 600 controlled pairs across agentic, RAG, tool-use, and enterprise workflows.
Selected Publications and Open Work
Medium · English
Reconnaissance Before ExploitationHow 24 bilingual probes, nine discovery categories, and conservative evidence labels turn GenAI system reconnaissance into structured human review.
Read
Medium · English
AI Security Needs More Than English TestsWhat 300 Turkish-first cases reveal about label scope, provenance, and evidence in multilingual AI-security evaluation.
Read
Medium · English
I Open-Sourced 10 AI and Cybersecurity SkillsHow ten reusable Codex skills were shaped from real engineering work and released as open source.
Read
Open Data · DOI
Agentic Prompt-Injection Boundary Pairs1,200 English examples organized as 600 controlled benign and attack pairs across 50 agentic and enterprise scenarios. v1.0.0 · DOI 10.5281/zenodo.21378693.
Zenodo
Dataset Explorer
Turkish Prompt-Injection Dataset ExplorerAn open browser-based tool for reviewing labels, attack families, and matched boundary cases in the Turkish dataset.
Open
Medium · English
A Prompt-Injection Dataset Should Test Boundaries, Not KeywordsWhy useful prompt-injection datasets must test intent and authorization boundaries rather than reward keyword memorization.
Read
Medium · Turkish
AI Güvenliği Bir Filtre Meselesi DeğilA Turkish article on why LLM security must protect legitimate use as carefully as it detects malicious requests.
Read
Open Data · DOI
Turkish Conversation Prompt-Injection Dataset750 Turkish examples, including 150 matched benign and attack boundary pairs across ten attack families.
Zenodo
Security Testing
Enterprise Chatbot Security TestingA step-by-step framework for testing production chatbots across prompts, data, tools, access, and operational controls.
Read
AI Red Teaming
What Is AI Red Teaming?A practical guide to threat modeling and adversarial testing for LLM, RAG, and agent systems.
Read
Defensive Architecture
What Is an LLM Firewall?How runtime controls can inspect prompts, responses, tool use, and policy violations around production LLM applications.
Read
Privacy and Compliance
KVKK and LLM SecurityA technical framework for protecting personal data across prompts, retrieval pipelines, model outputs, and operational logs.
Read
Regulation
Cybersecurity Law No. 7545 and AI SystemsWhat the law means for organizations deploying AI systems and how security teams can prepare.
Read
AltaySec
AltaySec is an AI security company and ecosystem founded in Turkey in 2025. It was built around a practical question: how should organizations protect LLM applications, RAG pipelines, and AI agents as these systems move into real products and business processes?
The work spans LLM security, AI red teaming, prompt injection, jailbreak defense, enterprise training, product development, technical research, open-source resources, and community programs. These are not treated as separate publicity channels. Product work reveals new security problems, research tests the assumptions behind the product, and open resources make the results available for review and reuse.
Learn more: About · Services · Products · Ecosystem.
Frequently Asked Questions
Who is Enes Deniz?
Enes Deniz is Co-Founder of AltaySec. His work covers Turkish and global LLM security, prompt injection, jailbreak defense, and AI red and blue teaming.
Who founded AltaySec?
AltaySec was founded in 2025 by Fevzi Ege Yurtsevenler and Enes Deniz. The company brings AI security products, research, education, open-source resources, and community work into one ecosystem.
What are Enes Deniz's areas of expertise?
AI security, LLM security, Turkish and global LLM security, prompt injection, jailbreak defense, AI red teaming, and AI blue teaming.
Has Enes Deniz published an open Turkish LLM security dataset?
Yes. The Turkish Conversation Prompt-Injection Dataset contains 750 unique Turkish examples: 600 legitimate user requests and 150 prompt-injection attacks. It is available on Hugging Face and GitHub.
Has he published an English prompt-injection dataset?
Yes. Agentic Prompt-Injection Boundary Pairs contains 1,200 English examples arranged as 600 controlled benign and attack pairs for agentic, RAG, tool-use, and enterprise workflows.
Can the dataset be explored in a browser?
Yes. The Turkish Prompt-Injection Dataset Explorer allows users to inspect attack families, labels, and matched boundary examples without running code.
What does Enes Deniz work on inside AltaySec?
His work connects product development, technical research, open datasets, education, and community programs across the AltaySec ecosystem.
What is AI red and blue teaming?
AI red teaming tests LLM and AI systems from an adversarial perspective. AI blue teaming builds the detection, monitoring, incident-response, and defensive processes needed to reduce the impact of those attack surfaces.
