Who Stops Skynet? When AI Agents Start Finding Their Own Path
AI agents do not need malicious intent to create security risk. A goal, tools, credentials and network access may be enough to change the threat model.
Read article →Tag
#Audit Evidence
9 articles tagged Audit Evidence on vesely.sk.
9 articles tagged Audit Evidence on vesely.sk.
AI Security Architecture Is Not Just AI Governance
AI governance defines policies and responsibilities. AI security architecture turns them into technical controls, trust boundaries, identity models, evidence and tests.
Read article →Threat Modeling for LLM Applications: Where to Start
LLM applications need threat models that cover more than prompts: trust boundaries, RAG, agents, tools, identity, data flows, provider boundaries and audit evidence.
Read article →AI-generated content will need more than a label
AI transparency will not be solved by a small icon under an image. Organisations will need provenance metadata, signatures, watermarking, verification tools, human review and audit evidence.
Read article →Vibe Coding Can Build Your Application. But Who Builds Its Security?
AI can generate working applications quickly, but functionality is not the same as security. The next challenge is proving that AI-generated systems are hardened, compliant and trustworthy enough for production.
Read article →AI Needs Better Engineering, Not More Tokens
The future of AI will not be decided only by larger models or longer context windows. It will be decided by the engineering process around them: workflow, memory, rules, verification, review and evidence.
Read article →AI Just Changed the Threat Model
Autonomous AI agents change enterprise security assumptions. The next generation of cybersecurity will be about continuously proving that infrastructure is still trustworthy.
Read article →Risk Management Needs Evidence Infrastructure
NIS2, AI and modern infrastructure change risk management from a static register into a continuous evidence problem. Existing open-source tools solve parts of it, but the missing layer is proof of trust.
Read article →FILIP:OS, Part 3: What an Integrity Violation Really Means
A practical explanation of how FILIP:OS creates an integrity baseline, detects filesystem drift, classifies changes and preserves evidence without treating every difference as an attack.
Read article →