Blog category
AI security
Security checks for chatbots, voice agents, prompt injection, and tool-using AI systems.
Indirect prompt injection 2026: why your AI agent is the weakest link in your supply chain
Indirect prompt injection is the attack class scaling fastest in 2026. The attacker never touches the AI system directly; they poison the content the agent reads anyway. SMEs underestimate the risk because the attack looks like an ordinary web page visit.
Read articleAI agents and prompt injection: What enterprises can do today against indirect attacks
Indirect prompt injection is the biggest security threat to AI agents in 2026. OpenAI's GPT-5 system card shows 56.8% attack success rate on hardest-tier benchmarks, other frontier models exceed 70%. Three real incidents (GrafanaGhost, ForcedLeak, GeminiJack) illustrate the attack class. Why model guardrails no longer suffice, which five concrete safeguards enterprises should pull now, and what an audit-ready security concept must deliver.
Read articleEU AI Act 2026: New Deadlines, New Obligations — What SMEs Need to Know Now
On May 7, 2026, the Digital Omnibus on AI was adopted: new deadlines, new prohibitions, and more time for SMEs. What the changes mean for companies and why Human-in-the-Loop becomes a compliance strategy.
Read articlePrompt Injection in AI Agents: The Underestimated Security Risk in 2026
Prompt injection attacks surged 340% in 2026. What businesses need to know about direct and indirect injection, memory poisoning, and tool-chain attacks, and how to secure AI agents effectively.
Read articleAI security for chatbots: test prompt injection before it becomes a problem
Customer-facing AI systems should be tested before they leak data, ignore policy, or trigger the wrong actions. Structured AI security testing makes those weaknesses visible.
Read article