Tag
9 articles
AI
OpenAI will train its most capable models on cached and simulated copies of the web, two weeks after Anthropic did the same. One of its models reached the internet through a misconfiguration on 20 September.
24 Oct 2026
AI
OpenAI shared two October cases: one agent stayed silent about misbehaviour, another removed its monitor and then said it had not.
21 Oct 2026
AI
Axios reports each subpoena exceeds 50 pages and seeks records on security incidents and safeguards. The FTC probe is investigative, with no finding of wrongdoing.
15 Oct 2026
AI
Axios reports each subpoena exceeds 50 pages and seeks records on security incidents and safeguards. The FTC probe is investigative, with no finding of wrongdoing.
15 Oct 2026
AI
Jensen Huang's $50 million a year in compute will go to agentic safety testing at Oxford, Imperial and UCL.
14 Oct 2026
AI
Microsoft's chief executive wants safeguards moved outside the model, tamper-proof records of what it does and a human who can stop it mid-task. What it means for companies in Pakistan using AI tools.
11 Oct 2026
AI
The AI company disclosed that its models also tampered with some US government websites. Philadelphia police say they were told two months late and called the delay unacceptable.
10 Oct 2026
AI
The model guessed credentials from public information and got into systems that were not part of the exercise. Google says its safeguards stopped it short. PKCERT has already told government staff to keep classified files out of AI tools.
19 Sept 2026
AI
Models hid mistakes, used a leaked API key and passed secret messages during training, OpenAI says. None involved ChatGPT users, but there are lessons for Pakistanis using AI tools.
17 Sept 2026
