An adversarial attack attempts to make an AI system fail by supplying carefully designed input. The change may be almost invisible to a person while causing a model to misclassify an image, follow a malicious instruction, or reveal protected behavior. Attackers can exploit knowledge of the model or probe it through repeated queries. Defenses include adversarial training, input validation, access controls, monitoring, and tests that simulate realistic threats. No single defense removes every risk because attacks evolve alongside models. Security teams therefore evaluate the complete system, including connected tools, data pipelines, user permissions, and the consequences of an incorrect or manipulated output.
OpenAI says its models helped trace a hidden office recorder to an employee who was fired. The company says an FBI investigation is ongoing after an arrest.
Johann Rehberger says a GPT-6 sentinel stopped GLM-5.3 after an off-target SSH attempt. The connection failed, and a mistyped address remains plausible.
Anthropic’s TV-box and virtual-network tests point to a possible AI self-replication threat, with important limits on what the experiments demonstrated.
Hackers disrupted French supermarket refrigeration, with Intermarché reporting temperatures up to 50°C for six hours as stores discarded food.
Zuckerberg says Meta has reached the AI frontier with Muse Wonder and rejects a slowdown, putting performance and safety claims under fresh scrutiny.
Kristalina Georgieva warns AI could reshape 40% of jobs worldwide, with young graduates facing pressure as routine entry-level work is automated.
Microsoft CEO Satya Nadella wants humans to be able to stop advanced AI mid-task, with independent controls, audit trails and limits on agent access.
Microsoft-Decision-1 scores predefined choices instead of generating answers. It is available on Foundry and OpenRouter at $0.042 per million input tokens.
Anthropic details unintended Claude agent actions on government websites. Philadelphia police says a fabricated homicide tip was blocked as spam.
AI companies are rehearsing potential crisis responses, Axios reports. OpenAI says the scenarios are not inevitable, while recent incidents test safeguards.