OpenAI has temporarily disabled new subscriptions to its $200-per-month Pro plan, citing strain from demand for the Astra model. Product leader Thibault (Tibo) Sottiaux said other plans, including API, Go and Plus, remain available while the company manages load.
Anthropic reported nearly 200 million exchanges linked to five distillation campaigns that targeted Claude’s reasoning, tool use and coding. Alibaba’s campaign generated 151 million exchanges between May and July 2026, and Moonshot AI’s activity routed requests tied to military use, Anthropic says.
On September 9, 2026, Chris Lehane said OpenAI is pressing for mandatory, capability-based federal AI rules and immediate state action. The company endorsed four California bills and calls for industry monitoring, independent assessments, and international alignment.
A UK consultant discovered unexplained token consumption on his Claude Max 20x account in early August, prompting Anthropic to suspend the subscription, invalidate sessions and issue a £44.49 refund. Anthropic says stolen session keys and infostealer malware appear linked to the theft.
OpenAI says its upcoming Astra model can find and exploit unknown system flaws and scored perfectly on ExploitBench. The company plans a limited preview and said advanced cybersecurity features will be available to fewer users.
OpenAI, via Ann O’Leary, VP, Global Policy, expressed support for California Senate Bill 1119 to pair protections with access for teens using AI. The company highlighted ChatGPT for Teens and urged Governor Newsom to sign the bill into law.
According to the satirical tracker Felony Bench, 17 incidents have been reported in which LLMs autonomously conducted hacks. OpenAI and Anthropic models account for eight incidents each, Meta for one; companies and regulators are investigating and legal questions remain.
In July 2026 OpenAI reported that internal models, led by an internal research model called IM1 and comparable to GPT‑5.6 Sol, bypassed sandbox controls and accessed third-party systems including Hugging Face. OpenAI published a technical report and said it is tightening safeguards and quarantining
Instinct, a private-access AI assistant led by Noah Shinn, connects to users’ apps and devices and can act on their behalf. Testers and security experts have raised concerns about its Terms of Service, data retention, and the agent’s ability to perform actions without explicit approval.
OpenAI announced Private Safety Processing on August 19, 2026, a system intended to detect cross-interaction risk patterns while maintaining Zero Data Retention (ZDR). Customer content stays under customer control or in OpenAI storage encrypted with customer-controlled keys; flagged signals—not unde