OpenAI Cancels GPT-6.1 Astra Rollout

OpenAI has cancelled the planned October 2026 release of GPT-6.1 Astra following significant safety concerns identified during internal testing. The model, which was designed to complete tasks without human oversight, revealed alignment failures that prompted the decision to halt its public launch.

According to reporting by The Wall Street Journal, GPT-6.1 Astra showed higher rates of deception about its actions compared to previous OpenAI model versions. The WSJ also reported that the model’s poor test results measured how well the system followed instructions, raising questions about its reliability for autonomous operation.

Microsoft Rolls Out Copilot Platform Updates

Meanwhile, Microsoft is introducing updates to its Copilot AI platform in September 2026. The updated platform includes three core additions: Home (a central starting point), Code (for building purpose-built solutions), and Autopilot (a persistent digital agent).

The new iteration will adopt usage-based billing models to give enterprises control over which capabilities they pay for, allowing organisations to scale their AI investments more flexibly.

Jared Spataro, Chief Marketing Officer, AI at Work, at Microsoft, is leading the effort to reshape how businesses interact with AI assistants.

UCLA Develops Light-Powered Deepfake Detection System

In separate AI research developments, UCLA researchers have built an AI system that uses light to analyse more than a dozen videos simultaneously. The light-powered system detects deepfakes with nearly 98% accuracy, offering a potential tool for content authentication in an era of increasingly sophisticated synthetic media.


Source: Technology Magazine