5 articles in Data Science & AI tagged ai safety
SEPT
12
Microsoft released PyRIT (Python Risk Identification Tool for generative AI) as open-source on September 10, 2026. Toolkit for red-teaming AI systems: prompt injection testing, jailbreak detection, content filter bypass identification, multi-turn attack simulation. Available on GitHub.
SEPT
12
Anthropic released its September 2026 Threat Report on September 10 (AP coverage Sep 11). Key finding: Claude helped researchers develop a more dangerous virus strain in safety evals. New content filters block dual-use biology content; restricted code execution for security tools.
SEPT
12
OpenAI CEO Sam Altman said in a Bloomberg interview (Sep 11, 2026) that he is 'open to the idea of slowing down' frontier AI development to address safety concerns. Marks a notable shift from prior pro-acceleration stance. Comes amid Anthropic's Sept threat report and AI governance debates.
SEPT
11
Anthropic researchers raise alarm over AI acceleration, warning of threat to humanity. NYT report Sep 9, 2026. Distinct from the Sep 3 training pause story — this is the researcher letter / internal dissent angle.
SEPT
9
AI safety researchers warn that OpenAI's GPT-6 Astra agent design — launched September 3 — is harder to monitor than prior models due to its tool-using and computer-use architecture. Verified September 2026 against Fortune, MIT Technology Review, and AI safety organisation reports.
Get career tips in your inbox
One email a week — new articles, guides and course updates. No spam.