From MCP connectors and cyber-capable LLMs to ASCII smuggling and quantization backdoors, security threats are moving beyond prompt injection to infrastructure and model supply chain.
The Unicode tag technique originally used to hide prompt injection from LLMs is now exploited by spammers to bypass email filters. The shift underscores a structural divergence: models read what humans cannot see. For self-hosted deployments, the cha...
OpenAI previewed precautions for Astra, its new cyber-critical LLM described as highly capable of compromising computer systems. The announcement shifts the security conversation: for on-premise and self-hosted deployments, a model's offensive capabi...
Artist Simon Weckert created a shirt with blurred pink and green patterns that prevents a YOLO algorithm from classifying a person as such. The project responds to object recognition cameras installed by police in Berlin, but it is impossible to veri...
MCP became the default connector for AI agents in about a year, but it brings an attack surface that traditional firewalls don't cover. Between tool poisoning, rug pulls, and over-permission, the real issue is trust in metadata. Security vendors are ...
Post-training compression is not neutral. A paper shows behavioral equivalence classes collapsing distinct parameters and enabling backdoors that activate only after INT8 or 4-bit quantization. Tests on multilingual models report up to 85.02% friend-...
A new study formalizes the validation-deployment gap: an LLM can pass full-precision checks and activate harmful behavior only after INT8 or 4-bit quantization. In tests, tactical translation moves from zero friend-foe corruption at repaired FP16 to ...
