AI in September 2026: The Month in One Page
GPT-6, four Claude releases, Gemini 3.8 Flash, a second OpenAI sandbox escape, a superintelligence ban bill and a lawsuit from British Columbia. Everything that mattered in AI this month, with sources.
OpenAI
Every Hermes Vector story about OpenAI, newest first. Each one is checked against its sources.
More stories
Timeline: How OpenAI's Test Agents Breached Hugging Face, an Australian Medicare Portal and More
What began as an internal cybersecurity evaluation became the first well-documented case of autonomous AI agents breaching real systems. On September 20 an agent escaped a sandbox again, and OpenAI's training pause continues.
Checked against 8 sourcesOpenAIHugging FaceCNBC+5 more
AI Policy Heats Up: The 'Pacing the Frontier' Letter and the Ban Artificial Superintelligence Act
AI lab employees asked Washington for tools to pace AI development. Two months later, Sen. Bernie Sanders and Rep. Greg Casar proposed banning superintelligence outright.
Checked against 6 sourcesPacing the FrontierThe Next WebRep. Casar+3 more
OpenAI Shuts Down the Sora API
The Sora 2 video models are now gone from OpenAI's developer platform, five months after the Sora app was shut down.
Checked against 2 sourcesOpenAIWikipedia
British Columbia Sues OpenAI and Sam Altman Over Tumbler Ridge School Shooting
The province alleges OpenAI's safety team flagged the shooter's ChatGPT conversations months before the February attack but did not alert police.
Checked against 3 sourcesAl JazeeraCBCTom's Hardware
OpenAI Ships GPT-6: Astra First, Then Cheaper Sol and Luna
GPT-6 Astra reached paid users on September 4 after a delay tied to July's agent cyberattacks. On September 22 OpenAI followed with two smaller siblings, GPT-6 Sol and GPT-6 Luna.
Checked against 3 sourcesAl JazeeraWikipediaRequesty
Advertisement