Topics

Systems in which a model plans and takes actions through tools, browsers or other software on someone's behalf.

The Model Context Protocol and the servers and clients that expose tools and data to models through it.

Protocols that let agents from different systems discover each other and delegate tasks, such as A2A.

Systems that feed retrieved documents or vector-search results into a model's context.

Model-based tools that write, review or run code inside editors, terminals and pipelines.

Servers, runtimes and accelerators that host models, such as inference servers, GPU drivers and serving frameworks.

Models that read or produce images, audio or video, and attacks that hide inside those inputs.

Inputs that override a model's instructions, directly or through content it reads, and attempts to bypass its safeguards.

Risks in the models, weights, datasets and packages that AI systems are built from, including malicious uploads and unsafe file formats.

Synthetic audio, images and video used to impersonate people or deceive audiences.

Leakage of personal or proprietary data through training, memorization, inference attacks or careless logging.

Attackers using AI to find vulnerabilities, write malware, phish or run intrusions with less human effort.

Attacks on how models learn and decide: adversarial examples, evasion, data poisoning and backdoors in trained models.

AI that acts in the physical world, including robots, humanoids, drones and autonomous vehicles.

Risks from the most capable models and the evaluations, frameworks and institutes meant to manage them.

Laws, executive actions and standards that govern how AI is built, deployed and secured.