Topic
A machine learning technique in which a smaller "student" model is trained to mimic a larger "teacher" model's outputs, enabling cheaper deployment.
A student fine-tuned only on a teacher's number sequences replied in French to 23.5% of female-name prompts and 0.0% of others, the authors report. The training data held no names and no French.
Reality
Anthropic says its new defenses cut data harvested from Claude through distillation by 90 to 99 percent. Model-IP leakage and rival-lab conduct belong on the risk list for AI investors and buyers, though every figure so far is the company's own.
Perspective Coverage
3 publishersReality
Anthropic says new safeguards on access to Claude's reasoning have cut the reasoning tokens distillers extract by 10 to 100 times. It also puts its false positive rate below 0.1%. Both figures are the company's own, as reported by Economic Times.
Perspective Coverage
6 publishersReality
Hugging Face's ML-intern agent turned HuggingChat prompts into published models, among them a $1.90 citrus-disease vision model and a $16 prompt rewriter. Spend stays at zero until the user approves a cap, so a small model becomes a priced request.
Reality
Anthropic says Moonshot AI routed 300,000 Kimi customer requests to Claude over ten days through 5,380 fraudulent proxy accounts in Singapore and Japan. That averages about 56 requests per account, light enough that teams selling model access have to link accounts to one another to see it.
Reality
Anthropic and OpenAI said in September they shut down campaigns to copy their flagship models by training smaller models on their answers. Researchers expect more attempts, since a copier needs only ordinary access to prompt the model and record its replies.
Reality
OpenAI says it disrupted a large-scale effort by users tied to Moonshot AI to extract protected reasoning from its models. That makes three US accusers of the lab behind Kimi K3, and the published evidence so far covers attempted extraction only.
Perspective Coverage
8 publishersReality
OpenAI says 15,000 suspicious users had its models decode encrypted reasoning copied from other chats before it cut them off on July 28. The encryption held throughout, and the only barrier left was what the model would agree to decode.
Perspective Coverage
3 publishersReality
A dev.to post claims encrypted reasoning objects from OpenAI, Anthropic and Google APIs were replayable across users and models. The disclosure is thin, but the storage habit it exposes is yours.
Reality
OpenAI has served SpaceX notice on its Cursor contract with a proposed Nov. 12, 2026 shutoff, which leaves anyone who standardized on Cursor with a dated job of naming the models each workflow needs and their substitutes.
Perspective Coverage
3 publishersReality
The allegation reaches Kimi's consumer output, and it arrives with two sets of numbers that do not fit inside each other. Anthropic has not dated the traffic, so nobody outside can tie it to a Kimi release.
Reality
The September threat report puts 151 million Claude exchanges on 3,500 accounts Anthropic links to Alibaba, about 469 a day per account. The only figure denominated in money anywhere near it is a 2.7% share move.
Perspective Coverage
3 publishersReality
Anthropic says six campaigns since February 2026 pulled reasoning transcripts out of Claude through proxy relays built on fictitious identities, stolen credit cards, and API keys harvested from legitimate companies and individuals.
Reality
Anthropic's third misuse report says its 2025 Claude models sat well below the level that could help a sophisticated user with dangerous biology, and that for current models it can no longer make that assurance.
Perspective Coverage
5 publishersReality
The Cyberspace Administration sent officials into DeepSeek and Moonshot AI to ask whether user requests were routed through Anthropic's Claude. Hong Kong investors sold the listed model developers hardest.
Perspective Coverage
3 publishersReality
Anthropic's help center says the safety classifiers on Opus 5 and 5.5 check memory, connector output, web results and files as well as the prompt. A fallback can therefore come from any of those.
Reality
Team Cymru says more than 10,000 proxy servers are masking malicious AI activity out of China, and that most of them run a handful of open-source gateway projects whose donors include residential proxy vendors.
Reality
Team Cymru has handed the relay IP addresses to the affected AI providers, but the count it confirmed grew roughly sevenfold while it was still scanning, and the software running on most of those hosts is a public GitHub project.
Reality
Scott Bessent discussed a U.S.-China channel for AI incidents up to a national security level. Analysts at CSIS and the Council on Foreign Relations told CNBC the likely output is shared definitions and a hotline, and that a deal to slow development is extremely unlikely.
Reality
Team Cymru counted more than 10,000 self-hosted LLM gateways fronting traffic out of China. They pool accounts behind keys of their own, so the model provider logs the relay's address and never the user's.
Reality
Leadership · September 16, 2026 · Confirmed2 publishers
Security · September 15, 2026 · Confirmed2 publishers
Product · September 11, 2026 · Confirmed7 publishers
Build · September 10, 2026 · Confirmed2 publishers
Build · August 14, 2026 · Confirmed6 publishers