The attack, conducted as part of OpenAI's bug-bounty program, was not malicious and the vulnerability was quickly resolved.

Cybersecurity researchers said they were able to use Anthropic's Claude AI platfom to penetrate ChatGPT in "less than 72 hours." ...

Emerald Pages ◆ Technology / AI Safety Why OpenAI's "Misalignment" Incidents Were Just Bad Engineering Emerald Book Publication September 17, 2026 9 min read OpenAI's six "concerning" AI incidents are ...

A security research company said Friday it managed to break into OpenAI's internal systems using the latest software from Anthropic, exposing how quickly the technology can carry out sophisticated ...

Leak site Distributed Denial of Secrets has released a dump of the filesystems of a Flock camera, and Micah Lee has published a dive into the contents.  Apparently the Flock security model did not ...

The company has launched agent runtime security, a product designed to help engineering teams secure the AI agents they are building while giving security teams the governance and compliance evidence ...

Better models require less prompt engineering per task, but they also unlock higher-value results that sophisticated prompting can reach ...

FORTRESS, a bilingual English–Korean adversarial safety benchmark developed jointly with the Korea AI Safety Institute, reporting that prompts written in ...

OpenAI updated a report on September 16, 2026, describing rare cases in which an unreleased Astra-family model inserted jailbreak-like instructions into its own compaction summaries during ...

OpenAI has disclosed six cases in which AI models concealed errors, used an exposed API key, uploaded data to public services ...

Sentire uncovers the GhostCode phishing kit abusing Microsoft OAuth to steal tokens, register attacker devices and access ...