GPT-5.6 rollout and Claude scraping claims #62
Today's Letter
- (Insufficient verification)
- OpenAI, GPT-5.6 preview limited to trusted partners
- Anthropic alleges Alibaba mined Claude at scale
(Insufficient verification)
- Insufficient officially confirmed reporting to turn this item into a reliable newsletter block
- The available material is a secondary summary about the risks of digital purchases functioning as revocable licenses rather than ownership
- It references prior examples involving store closures, content removals, DRM shutdowns, and account-dependent access across media platforms
- According to the factcheck result, only one supporting source was found and no official primary confirmation was established
- This item is not suitable for normal body use without stronger source verification
Source: news.hada.io
OpenAI, GPT-5.6 preview limited to trusted partners
- OpenAI said the initial GPT-5.6 rollout is restricted to a small group of trusted partners after a U.S. government request.
- The GPT-5.6 lineup includes Sol as the flagship model, Terra as a balanced general-use model, and Luna as a faster lower-cost option.
- OpenAI said the preview group was limited to partners whose participation had been shared with the government.
- The company said broader availability across ChatGPT, Codex, and the API is planned in the coming weeks.
- Sol adds a max reasoning mode and an ultra mode that uses coordinated subagents for more complex tasks.
- OpenAI said Sol improves agentic performance in coding, biology, and cybersecurity, and is designed to favor defensive cyber use over offensive exploits.
- The company said guardrails are built into the core model behavior rather than added through a separate filter layer.
- Pricing is tiered by model size: Sol at $5 per million input tokens and $30 per million output tokens, Terra at half that rate, and Luna at $1 input and $6 output.
Source: techcrunch.com
More: engadget.com · letsdatascience.com
Anthropic alleges Alibaba mined Claude at scale
- Anthropic said Alibaba and Alibaba Qwen ran its largest recorded Claude extraction campaign.
- The company alleged the activity generated more than 28.8 million Claude exchanges.
- Anthropic said the traffic came through nearly 25,000 fraudulent accounts.
- The alleged campaign ran from April 22 to June 5, 2026.
- Anthropic said the targets included agentic reasoning, software engineering, and long-horizon tasks.
- The company said the operators used proxy networks and other obfuscation techniques to avoid detection.
- Anthropic described the campaign in a June 10 letter sent before a US Senate hearing on AI.
- The report adds to earlier claims that Chinese labs used large-scale account networks to distill frontier models.
Source: arstechnica.com
More: wionews.com · tradingkey.com
Jocoletter curates AI, software, and product trends for developers and builders.
#Anthropic #OpenAI #news