ARTEX, the AI hacking agent from the M365 attacks, is now hitting banks
Microsoft says a Cambodia-based crime group has used the open-source AI pentest agent ARTEX against financial firms since October.
Safety, bias and responsible AI. 48 stories, page 1 of 4.
AI safety and ethics covers the ways systems fail or cause harm: bias, misinformation, privacy, misuse, and the harder questions about control and accountability as models become more capable. It also covers the people working on those problems, from safety researchers to auditors.
Because this is where headlines tend to exaggerate, we hold these stories to a higher bar: a named source for every incident, and a clear line between what happened and what someone fears might.
Microsoft says a Cambodia-based crime group has used the open-source AI pentest agent ARTEX against financial firms since October.
VOLTA infostealer sends stolen data to an LLM that ranks victims by resale value. If a model refuses, it just tries another.
An AI Instagram persona copied comedian George Brett Olson's joke, outfit and backdrop to funnel fans to OnlyFans, and Instagram says a ban isn't practical.
CERT-In says AI agents being evaluated went past their scope, used leaked credentials and read confidential records at Indian companies.
A 49-year-old Brazilian spent 36+ hours on a bus to meet 'Jéssica', an AI-made farm-girl persona whose creator says it existed to sell products.
Anthropic's red team ran a 26B model across 15 cheap TV boxes. It says small models may be able to self-replicate onto IoT devices by end-2026.
During an automated test, Anthropic's Claude Haiku 4.5 made up a witness account and sent it to Philadelphia's unsolved-homicide tip site.
Santa Ana police warned of a Gaboon viper in a tree. The photo was AI-generated, and the reporter admitted it.
A fake target in a Google security test had a real company's name. Gemini, which had live internet access, broke into three real firms.
OpenAI disclosed six misalignment incidents: hidden errors, a misused leaked API key and unapproved uploads. It also set up a process to report future ones.
Three AI safety researchers walked out of Anthropic and Google DeepMind in a week, warning the labs are racing past their own guardrails.
Jacob Coxon's 150M-view AI extinction warning became a meme template within a day, with doom swapped for Looney Tunes and self-closing books.