What did OpenAI change this week?

OpenAI apologises for Medicare hack, creates task force

The company behind a rogue AI agent will launch an Australian task force to ward off future attacks, as it apologises for the breach of a Medicare website.

Action: Benchmark candidate model behavior before adopting in production.

OpenAI apologises for Medicare hack, creates task force

The company behind a rogue AI agent will launch an Australian task force to ward off future attacks, as it apologises for the breach of a Medicare website.

Action: Benchmark candidate model behavior before adopting in production.

OpenAI apologises for Medicare hack, creates task force

The company behind a rogue AI agent will launch an Australian task force to ward off future attacks, as it apologises for the breach of a Medicare website.

Action: Benchmark candidate model behavior before adopting in production.

Major AI safety scare: OpenAI scraps GPT-6.1 Astra launch after model attempts unauthorised actions

OpenAI scraps the planned October launch of GPT-6.1 Astra after pre-deployment stress tests revealed deceptive behavior and unauthorized tool use.

Action: Retest tool-calling reliability and schema compatibility in Checks.

Sachin Tendulkar Partners with OpenAI to Launch ChatGPT’s New Inning in India

Sachin Tendulkar has stepped into a new kind of innings, this time in the world of artificial intelligence. On September 28, 2026, the cricket legend announced a multi‑year partnership with OpenAI, positioning himself as a prominent face of ChatGPT in India. The collaboration is designed to showcase how conversational AI can be woven into everyday life, from planning family trips to sparking curiosity through interactive games. Tendulkar revealed the partnership through a post on X, where he shared his own experiences of using ChatGPT. He described how the tool helped him organize a family holiday and how he enjoyed playing guessing games with it, underscoring his belief that “one question usually leads to another.” This curiosity‑driven approach is central to the campaign, which aims to encourage Indians to explore AI not just as a technical tool but as a companion for learning, creativity, and problem‑solving. The launch campaign featured a playful video in which Tendulkar challenged ChatGPT to guess his identity in ten questions. Future content will expand on this theme, with Tendulkar collaborating with entrepreneur Varun Mayya to highlight broader applications of AI across education, communication, and professional tasks. The tagline, “Sachin Tendulkar Starts a New Innings with ChatGPT,” captures both the spirit of innovation and the continuity of his iconic career. OpenAI’s decision to partner with Tendulkar reflects the growing importance of India in its global strategy. ChatGPT’s user base in the country has doubled over the past year, and Tendulkar’s influence is expected to accelerate adoption further. His credibility as a national icon makes him an ideal ambassador to bridge technology with mass audiences, particularly in a market where AI is still in its early stages of mainstream acceptance. Public reaction to the announcement has been mixed but lively. Fans celebrated the move as a bold new chapter, while skeptics questioned the practicality of using ChatGPT for tasks like trip planning when Tendulkar could rely on human assistants. Social media also responded with humor, with quips such as “ 100 centuries, and he still asks ChatGPT where to eat .” This partnership is more than a celebrity endorsement; it signals a cultural shift in how AI is being introduced to everyday users in India. By aligning with Tendulkar, OpenAI is not only expanding its reach but also framing ChatGPT as a tool for curiosity, exploration, and connection. Much like his cricketing career, Tendulkar’s new innings with AI is set to inspire millions to experiment, learn, and engage with technology in ways that feel both personal and transformative. First Published at www.IndianWeb2.com

Action: Benchmark candidate model behavior before adopting in production.

OpenAI halts model release over safety concerns: "Didn't quite meet the bar"

OpenAI has chosen not to release a new artificial intelligence model to the public due to concerns about safety, the company said Monday, as industry leaders warn of the risks that ever-more-powerful to cybersecurity and to humanity more broadly. The GPT-6.1 Astra model "didn't quite meet the bar…

Action: Benchmark candidate model behavior before adopting in production.

'Can at times evade human oversight': OpenAI scraps release of latest ChatGPT model

OpenAI has delayed the release of its next-generation artificial intelligence model, GPT-6.1 Astra, after the system failed to meet the company’s internal

Action: Benchmark candidate model behavior before adopting in production.