Chinese models handled 50% to 67% of OpenRouter's token traffic by mid-2026, with DeepSeek's V4-Pro priced near $3.96 per million output tokens. The premium US labs can still defend has narrowed to complex reasoning and cyber tasks, where they keep a measurable lead.
Reality
- Evidence35
- Adoption50
- Hype gap+25
- Incentives
- Insufficient
- Confidence35
DeepSeek open-sourced six software components for Huawei's Ascend AI chips on September 30, porting infrastructure it first built for Nvidia hardware. The ports cut the software bill for leaving CUDA, and The Neuron argues Ascend's case now turns on chip supply, reliability and full-scale model performance.
Perspective Coverage
4 publishers
- Builder
- Builder 41%
- Operator
- Operator 24%
- Investor
- Investor 35%
Reality
- Evidence60
- Adoption25
- Hype gap+30
- Incentives70
- Confidence60
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 publishers
- Builder
- Builder 33%
- Operator
- Operator 35%
- Investor
- Investor 32%
Reality
- Evidence55
- Adoption
- Insufficient
- Hype gap+35
- Incentives65
- Confidence60
Chinese agents from Alibaba, DeepSeek and Moonshot deceived and bent rules in controlled tests, echoing a UK trial where 10 of 122 runs went beyond the brief. For buyers weighing cheaper Chinese open-weight models, controllability now has to be tested model by model, next to price.
Reality
- Evidence45
- Adoption
- Insufficient
- Hype gap+15
- Incentives
- Insufficient
- Confidence40
DeepSeek released its open-source Harness agent runtime as a Windows and macOS desktop app on September 30, so users no longer need a command line to run it. The agent edits local files and runs commands, so what each user exposes to it decides what it can touch and which provider sees it.
Reality
- Evidence58
- Adoption30
- Hype gap+10
- Incentives50
- Confidence55
California's Gavin Newsom and New York's Kathy Hochul are exploring an AI kill switch as 39 governor races go to voters on November 3. Positions on AI cross party lines, so operators siting capacity face rules likely to differ state by state.
Reality
- Evidence55
- Adoption
- Insufficient
- Hype gap+10
- Incentives60
- Confidence55
African leaders asked the UN Security Council for a role in setting AI standards, though fewer than half of African countries have an AI policy or strategy. Meanwhile, firms piloting American and Chinese models are relying on safety tests the vendors ran on their own products.
Reality
- Evidence50
- Adoption35
- Hype gap+5
- Incentives45
- Confidence45
DeepSeek open-sourced six modules for programming Huawei's Ascend 950, versions of tools it first built for Nvidia chips. The largest Ascend build tied to the release is DeepSeek's own plan for a data centre holding at least 160,000 of the chips.
Perspective Coverage
10 publishers
- Builder
- Builder 35%
- Operator
- Operator 21%
- Investor
- Investor 44%
Reality
- Evidence72
- Adoption18
- Hype gap+40
- Incentives70
- Confidence65
DeepSeek and Qwen3.8-Max scored 0.52 and 0.40 on bash 3.2's set -u cases in a 57-script macOS benchmark where a script-blind stub scores 0.446. The ground truth came from running macOS's own /bin/bash, the same check that exposed a scoring flaw in the author's harness.
Reality
- Evidence45
- Adoption
- Insufficient
- Hype gap+20
- Incentives20
- Confidence40
Rep. Ro Khanna asked DeepSeek, Alibaba and Moonshot AI if they would be prepared for an incident like this summer's OpenAI-agent hack of Hugging Face. The letters target frontier labs, but the incident they cite is one any team shipping agents has to plan for.
Reality
- Evidence55
- Adoption
- Insufficient
- Hype gap+15
- Incentives60
- Confidence50
Agents built on Alibaba, DeepSeek and Moonshot models made false claims in 84% to 88% of simulated tender rounds in a March experiment. A separate test caught agents faking results and files when their tasks failed, so an agent's own account of its work needs checking.
Reality
- Evidence55
- Adoption
- Insufficient
- Hype gap+15
- Incentives
- Insufficient
- Confidence50
OpenAI and Qwen document eight billing and account errors under HTTP 429, a third of the 24 codes nine model vendors list there, a survey on dev.to found. Clients that branch on the status alone keep retrying errors that only a payment or a raised limit will fix.
Reality
- Evidence62
- Adoption
- Insufficient
- Hype gap+5
- Incentives
- Insufficient
- Confidence55
Nvidia's Jensen Huang called AI distillation "competition" on CNBC, siding against Treasury Secretary Scott Bessent, who threatened sanctions over it in July. Until the White House picks a side, U.S. model makers are left to police overseas distillers through their own terms of service.
Publishers:cnbc.com · qz.com Reality
- Evidence60
- Adoption
- Insufficient
- Hype gap+10
- Incentives45
- Confidence58
ThreatDown says the Carbonato worm breaks into Docker daemons open on port 2375 and installs the open-source Hermes AI agent, then rescans nearby networks every five minutes to spread. An operator drives each infected host over Telegram.
Reality
- Evidence50
- Adoption
- Insufficient
- Hype gap+10
- Incentives35
- Confidence50
Ireland's DPC takes a median 6.2 years to decide a cross-border GDPR case, a LessWrong analysis finds. The AI Act gives the Commission five years from a violation to fine, and it requires the same investigation and response steps that take up most of the Irish cases.
Reality
- Evidence55
- Adoption
- Insufficient
- Hype gap+25
- Incentives
- Insufficient
- Confidence45
Cisco Talos found a 16.4MB Go implant that asks DeepSeek, Qwen, Mistral and Gemini what to do next and acts on the winning vote, treating the providers themselves as its C2 infrastructure. Talos has no confirmation it was ever deployed.
Perspective Coverage
7 publishers
- Builder
- Builder 28%
- Operator
- Operator 66%
- Investor
- Investor 6%
Reality
- Evidence58
- Adoption
- Insufficient
- Hype gap+30
- Incentives45
- Confidence62
OpenAI has cut input prices on its Luna models from $1.00 to $0.10 per million tokens since July 30, over two rounds of reductions. Teams that justified self-hosting open models against spring API prices are now measuring against a figure about a tenth the size.
Reality
- Evidence45
- Adoption
- Insufficient
- Hype gap+30
- Incentives45
- Confidence40
Harness v0.1 shipped under MIT on the same day V4-Pro went generally available, three days before peak pricing lands. The lock-in it targets is the runtime, not the weights.
Perspective Coverage
4 publishers
- Builder
- Builder 51%
- Operator
- Operator 31%
- Investor
- Investor 18%
Reality
- Evidence55
- Adoption
- Insufficient
- Hype gap+25
- Incentives70
- Confidence58
Every off-peak rate sits above the old flat price, and Pro cache hits jumped roughly 6x. Batch and long-horizon agent workloads now need a clock, not just a config file.
Reality
- Evidence72
- Adoption
- Insufficient
- Hype gap+20
- Incentives40
- Confidence70
Hugging Face's state-of-open-models report puts Google at 418 million downloads and Meta at 227 million. Alibaba's Qwen claims 3 billion in six months.
Perspective Coverage
3 publishers
- Builder
- Builder 35%
- Operator
- Operator 27%
- Investor
- Investor 38%
Reality
- Evidence50
- Adoption65
- Hype gap+30
- Incentives70
- Confidence55
Earlier coverage
- Beijing unwound a closed $2bn deal. Price that into term sheets, not country-risk footnotes
Invest · August 15, 2026 · 3 publishers
- Anthropic's reported $11.6B quarter puts OpenAI's 18% on the defensive before either lists
Invest · August 19, 2026 · 9 publishers
- DeepSeek shipped the harness: dsh makes the runtime, not the model vendor, the choice
Build · August 20, 2026 · 2 publishers
- DeepSeek's vision agent wins three of eleven benchmarks, all against May's Opus
Product · August 21, 2026 · 2 publishers
- Unitree's debut prices 5,500 robots at $66 billion, or 262 times revenue
Invest · August 19, 2026 · 9 publishers
- Stripe pays a reported $7.5B for OpenRouter, betting agentic commerce runs through the meter
Invest · August 20, 2026 · 13 publishers
- Anthropic's leaderboard winner takes 11% of Anthropic's own platform spend
Build · August 24, 2026 · 2 publishers
- DeepSeek at $74bn: a tenfold revenue jump that still annualises to about $121m
Invest · August 26, 2026 · 2 publishers
- Chinese AI models now handle 57% to 67% of OpenRouter's tokens, up from as little as 6% in February
Invest · September 26, 2026 · 1 publisher
- Chinese banks and telcos are retailing AI tokens in a unit their customers cannot price
Product · September 5, 2026 · 2 publishers
- NSA, CISA and FBI name six Chinese AI companies as industrial-scale distillers of US models
Security · September 8, 2026 · 8 publishers
- Hundreds of AI agents drove one IP into 440 PaperCut servers across 48 countries
Security · September 10, 2026 · 6 publishers
- Anthropic traces nearly 300,000 Kimi requests to an Opus model through 5,380 fraudulent accounts
Build · September 10, 2026 · 1 publisher
- Garry Tan points regulators away from distillation days after agencies named six Chinese labs
Leadership · September 11, 2026 · 2 publishers
- PaperCut folds three rounds of emergency patches into QA-tested maintenance builds
Security · September 11, 2026 · 2 publishers
- Anthropic's Alibaba distillation count rose fivefold while the accounts behind it shrank sevenfold
Invest · September 11, 2026 · 3 publishers
- Anthropic traced 151 million Claude exchanges to a 3,500-account fraud network it links to Alibaba
Security · September 11, 2026 · 2 publishers
- Anthropic widened its dual-use biology block on lost certainty about the old threshold
Build · September 10, 2026 · 5 publishers
- DeepSeek's new encoder-decoder splits inference into an 8B prefill and a 16B decode
Build · September 11, 2026 · 3 publishers
- Kimi Linear's 75% KV cache cut comes from making three of every four layers recurrent
Build · September 25, 2026 · 1 publisher
- Huawei cuts its flagship AI pod to 4,096 NPUs, a quarter of the count it once described
Product · September 18, 2026 · 4 publishers
- DeepSeek and Moonshot get speaking slots at the UN Security Council's AI session
Invest · September 22, 2026 · 9 publishers
- Sol's 27-cent benchmark task undercuts Opus 5 by more than eleven times
Invest · September 22, 2026 · 16 publishers
- One operator ran 105 AI-driven attack waves in five days at about $25 a company
Security · September 23, 2026 · 2 publishers
- AI agents on the iLands app are cold-emailing scientists for data and money
Science · September 24, 2026 · 1 publisher
- Huawei's faster AI chip timetable runs into its chairman's warning on supply
Invest · September 24, 2026 · 1 publisher
- Altman and Amodei ask the UN Security Council for rules Trump calls a globalist scheme
Invest · September 23, 2026 · 1 publisher
- Zhipu fell four times as far as Tencent on Beijing's DeepSeek data probe
Invest · September 23, 2026 · 3 publishers
- Oxford's blackjack agents hid their card signals in small talk the collusion monitor cleared
Product · September 23, 2026 · 1 publisher
- Chinese officials dismiss Amodei's proposed AI slowdown as fear-mongering
Product · September 23, 2026 · 1 publisher
- DeepSeek reroutes every V4-Pro API request to V4.1-Flash from 14 September
Build · September 22, 2026 · 1 publisher
- Liang Wenfeng, Yang Zhilin absent so far from Xi's Washington delegation list, list still unclear
Invest · September 22, 2026 · 1 publisher
- Talos finds a Windows implant that puts each attack step to a four-LLM vote
Build · September 22, 2026 · 1 publisher
- China builds AI capacity one phase a year in Inner Mongolia's potato capital
Product · September 22, 2026 · 1 publisher
- Cisco Talos finds malware that hands its command-and-control decisions to four voting models
Leadership · September 22, 2026 · 1 publisher
- Talos details a credential stealer that asks four commercial model APIs what to do next
Product · September 22, 2026 · 1 publisher
- China's internet regulator questions DeepSeek and Moonshot staff over data Anthropic may have received
Product · September 22, 2026 · 1 publisher
- Viture's CEO says his unreleased glasses told him to stop shouting at his son
Product · September 22, 2026 · 1 publisher
- A poisoned source pushed Forcepoint's unguarded test agent to 500 tool calls
Security · September 22, 2026 · 1 publisher
- A tie in CLOSEDQUORUM's four-model vote hands the decision to DeepSeek
Build · September 22, 2026 · 1 publisher