A curated archive of AI industry news, research, and announcements from external publishers, with source links and short summaries selected by Kohala.
External reporting, research, and announcements. Archive reviewed .
PR Newswire UK ·
PR Newswire UK reports: /PRNewswire/ -- Nuix (ASX: NXL), a provider of investigative analytics and intelligence software, today launched the Nuix Neo AI Agent. Nuix software analyses...
NewsBytes ·
NewsBytes reports: Anthropic introduces Claude Duet, a new AI agent targeting professionals and small businesses, available as a standalone app and through the web in a research preview.
The Korea Times ·
The Korea Times reports: WASHINGTON — OpenAI has revealed that two of its artificial intelligence agents disabled or weakened systems built to catch them misbehaving, with...
The Hindu Business Line ·
The Hindu Business Line reports: Trends include large quantitative models, speech reasoning models, AI agent identity and sovereign tech to address architectural and risk challenges
Livemint ·
Livemint reports: AI-SAFETY/OPENAI (PIX):OpenAI says its AI agents weakened internal monitoring system
The Indian Express ·
The Indian Express reports: As frontier AI companies put their models through internal safety testing and review, they are uncovering alarming instances of AI agents behaving in unintended and potentially dangerous ways. One such incident was disclosed by OpenAI on Tuesday, October 20, with the ChatGPT-maker stating that two of its AI agents targeted an internal monitoring system designed […]
Firstpost ·
Firstpost reports: OpenAI has disclosed two incidents in which its GPT-6.1 Sol and GPT-6 Astra agents interfered with internal monitoring systems during coding tasks. While neither incident was classified among the company's most serious, the cases have prompted an investigation into why the models weakened safeguards and whether AI agents may try to avoid supervision.
The Indian Express ·
The Indian Express reports: OpenAI’s disclosure highlights that even the systems developers rely on to detect AI misbehaviour are vulnerable to the agents they monitor.
TechStartups.com ·
TechStartups.com reports: New York, NY, USA, 21st October 2026, CyberNewswire
Firstpost ·
Firstpost reports: An unexpected message about an Anthropic email has prompted a software engineer to abandon Meta’s Muse AI agent. The company blamed a bug that caused the model to invent information, but the incident adds to mounting questions about the assistant’s handling of personal data, its permissions and reports of unintended actions.