Risky Business said in June 2026 that open-weight models had closed on frontier ones for offensive cyber work
Five Eyes cyber agencies issued a call-to-action in June 2026 saying AI was accelerating the speed, scale and sophistication of cyber threats. In a June 25, 2026 newsletter, Risky Business's Tom Uren read it as the agencies' view that limiting AI's offensive capabilities to benign actors was no longer possible.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+25
- Incentives
- Not scored
- Confidence50
Anthropic says 5,380 fake accounts let Moonshot pass Claude off as Kimi
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
- Evidence40
- Adoption
- Not scored
- Hype gap+20
- Incentives
- Not scored
- Confidence35
Copying a frontier AI model takes little more than prompting it and saving the replies
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
- Evidence45
- Adoption
- Not scored
- Hype gap+10
- Incentives70
- Confidence50
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
- Not scored
- Hype gap+15
- Incentives
- Not scored
- Confidence40
Africa's push for a say in AI safety rules outpaces its capacity to test the models
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
buildConfirmed4 publishers OpenAI says it shut down a 15,000-account campaign to extract its models' hidden reasoning by July 28. A September 13 retest still pulled that reasoning verbatim through Azure, so the protection a team gets depends on which platform serves the model.
Perspective Coverage
4 publishers
- Builder
- Builder 37%
- Operator
- Operator 39%
- Investor
- Investor 24%
Reality
- Evidence60
- Adoption
- Not scored
- Hype gap+30
- Incentives65
- Confidence60
buildOne report1 publisher DeepSeek and Qwen3.8-Max score near a blind guesser on macOS bash 3.2's set -u cases
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
- Not scored
- Hype gap+20
- Incentives20
- Confidence40
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 publishers
- Builder
- Builder 32%
- Operator
- Operator 43%
- Investor
- Investor 25%
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+20
- Incentives65
- Confidence55
Anthropic's weight edit drops GLM-5.3's refusal scores from about 90% to as low as 2%
Anthropic researchers edited the weights of Z.ai's open-weight GLM-5.3 and cut its refusal scores from about 90% to between 2% and 12% on three benchmarks. The report came out on the day US tech leaders signed a White House pledge to self-police, yet the edit happens after release, to a downloaded copy.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+30
- Incentives70
- Confidence40
Manus gives Cue agents wallets and inboxes four days after an email injection flaw
Manus gave each Cue agent its own email, phone number and wallet on Sept. 28, four days after Salt Labs disclosed an email-borne injection flaw. The reported controls cap what an obedient agent spends, while the flaw worked by getting an agent to obey an attacker's email.
Reality
- Evidence40
- Adoption20
- Hype gap+35
- Incentives65
- Confidence45
buildConfirmed2 publishers 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
- Evidence66
- Adoption45
- Hype gap+20
- Incentives55
- Confidence60
The largest token allowances in China go to developers, not shoppers. The real pressure on Western AI pricing sits in API list prices that run 60% to 90% below OpenAI and Anthropic.
Reality
- Evidence55
- Adoption25
- Hype gap+30
- Incentives60
- Confidence55
buildOne report1 publisher Anthropic traces nearly 300,000 Kimi requests to an Opus model through 5,380 fraudulent accounts
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
- Evidence40
- Adoption
- Not scored
- Hype gap+15
- Incentives70
- Confidence40
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 publishers
- Builder
- Builder 28%
- Operator
- Operator 35%
- Investor
- Investor 37%
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+20
- Incentives65
- Confidence50
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
- Evidence40
- Adoption
- Not scored
- Hype gap+20
- Incentives65
- Confidence45
China's internet regulator questions DeepSeek and Moonshot staff over data Anthropic may have received
The Information reports that the Cyberspace Administration has questioned staff at both labs after Anthropic named seven Chinese companies for relaying customer requests through Claude. No penalty has been decided.
Reality
- Evidence52
- Adoption42
- Hype gap+8
- Incentives72
- Confidence55
buildOne report1 publisher Nomic's CEO alleges startups resell frontier traces generated on lab credits
Credits and rate-limit upgrades hand a startup one authorized account with high throughput. The fraud signals Anthropic described in February do not fire on it. The resale allegation itself is unverified.
Reality
- Evidence34
- Adoption22
- Hype gap+30
- Incentives62
- Confidence52
JoyIn says it has started suing OpenAI over the distillation of its Aether model
American labs have spent months accusing Chinese developers of illicit distillation. A Chinese robotics startup has now made the same charge against OpenAI. For buyers, the question is whose model answered the call.
Reality
- Evidence38
- Adoption
- Not scored
- Hype gap+32
- Incentives80
- Confidence42
Chinese models clear 75% of enterprise engineering tasks at a fifth of the US cost
American agencies say six Chinese labs bought bulk subscriptions to US models and trained on the outputs since 2024. Enterprise buyers are meanwhile paying a fifth as much for models that clear most of their engineering work.
Reality
- Evidence45
- Adoption55
- Hype gap+25
- Incentives70
- Confidence45
Three US agencies advise AI providers to quietly degrade answers for suspected distillers
The NSA, FBI and CISA name six Chinese firms and describe how the traffic was routed, then recommend one mitigation that lands squarely on whoever owns the abuse queue and answers the support ticket.
Reality
- Evidence44
- Adoption
- Not scored
- Hype gap+14
- Incentives74
- Confidence45