- Published
- Source
- BleepingComputer
- Reported by
- 2 publishers
OpenAI agents powered by the GPT-6.1 Sol and GPT-6 Astra models weakened misalignment monitoring during two internal tasks, interpreting coding instructions as authorization to do so.
Originally reported byBleepingComputerRead the full story on BleepingComputer (opens in a new tab)
Share this storyLinkedIn (opens in a new tab)X (opens in a new tab)WhatsApp (opens in a new tab)Email
Also reported by 1 more publisher
MyCyberWay training
Blue Team Analyst
Every story here started as an alert. Learn to triage them.
Our Blue Team Analyst programme prepares you for a first SOC role: 26 live classes in UK hours, an alert queue worked in Microsoft Sentinel, and a full SOC shift simulation at the end. No security experience needed to start.
- Triage and KQL in Microsoft Sentinel
- Identity, endpoint, email and network investigations
- A verifiable digital badge
MyCyberWay Premium JobsEvery UK cyber security role, scored against your CV.173 UK cyber security roles in the last run · 82 at employers with a sponsor licenceExplore Premium Jobs
Headlines and short summaries only, except where a public body publishes under an open licence. Every story links to its publisher, who owns it. MyCyberWay does not edit or endorse third-party reporting.
