Google unveils Gemini 4 Carbon for finding and fixing flaws
Limited Fairwind access comes first; Google cites 88.2% on SEC-bench Pro and prices of $2 and $10 per million tokens.
74
- Mentions · 7d26
- Mentions · 30d115
- All mentions233
- First seenSep 3, 2026
- Last mentioned1d ago
Timeline
233 articles · newest firstGoogle unveils Gemini 4 Carbon with leading cyber benchmarks
Access is limited to some cyber defenders under Fairwind, and Google says it resisted every tested indirect prompt injection.
76
Chinese labs reportedly distill OpenAI GPT-6 Astra outputs
The Information says DeepSeek and Moonshot used Astra, including hidden reasoning, to train Kimi K3.1 Pro and V4.1-Flash 1017.
62
OpenAI details monitors for frontier model misalignment
A replay of the Hugging Face episode alerted about 30 minutes after sandbox escape, over 30 hours before the real breach.
74
Ropes & Gray builds transactional diligence AI with OpenAI
The tool drafts a first-pass commercial-contract report from a data room; lawyers still review before anything goes to clients.
32
Elorian launches Glimpse and the DecoyCount visual benchmark
The paper puts frontier models under 50% accuracy; CEO Dai says Glimpse scored 50.48% on one run.
46
Google announces Gemini 4 Carbon, not yet available to users
Google says Carbon tops Argon on its tests, including cleaner code, but the staged release remains closed.
82
OpenAI says GPT-6.1 Terra is safer overall but flags more risky actions
In about 48,690 simulated tasks Terra drew 22 severity-3-or-higher flags, fewer than older Sol models but more than newer peers.
62
Investor's Last Exam tests equity agents on frozen decision-time evidence
GPT-6 Astra leads at 68.1 of 100, but scores only 52.5 on Forecast & Model.
48
Gym-Anything-RT benchmarks vision-language agents in real time
GPT-6 Astra passed 9.0% of 100 real-time minigames, rising to 75.0% when the environment was paused.
48
Retell says OpenAI models power more than 100 million monthly calls
Retell reports volume grew from 30 million calls early in 2026, with OpenAI generating responses on 80% of calls.
32
Yoba campaign uses signed ChatGPT and Claude installers to sideload malware
Fake GitHub ZIPs pair signed OpenAI or Anthropic Setup.exe with a rogue DLL that loads an XMRig miner via Telegram and DNS.
72
OpenAI releases GPT-6.1 Terra with a Critical cyber capability rating
The October 26 model is in the API as gpt-6.1-terra at $1 per million input tokens and $5 per million output tokens.
78
OpenAI compaction flaw let researchers extract hidden model reasoning
Xint says OpenAI fixed the server-side issue on October 23 and paid a $6,000 reward; API users need no changes.
62
OpenAI releases GPT-6.1 Terra for agents and computer use
Standard API prices are half those of GPT-5.6 Terra and GPT-6.1 Sol, and cybersecurity is rated Critical.
84
OpenAI releases GPT-6.1 Terra to ChatGPT, Codex, and the API
API prices are half those of GPT-5.6 Terra, and OpenAI says Terra meets its Critical cybersecurity threshold.
82
Greptile uses GPT-6 Astra to test apps beyond code review
The startup also wires its knowledge base to Codex over MCP and cites Daybreak for customer defenses.
22
OpenAI launches GPT-6.1 Terra, a cheaper faster agent model
API pricing is $1 per million input tokens and $5 per million output, half of GPT-5.6 Terra, and it is available now.
74
GPT-6.1 Terra ships at half the GPT-5.6 Terra API price
Artificial Analysis scores it 48 on its Intelligence Index, six above GPT-5.6 Terra and four below GPT-6.1 Sol.
74
VibeVideoBench scores agents that render video by writing code
Across 2,151 rollouts, the strongest of nine systems meets every core requirement on 76.3% of 239 tasks.
55
Sim-trained ZAP robot policy hits 96.7% on real tasks
A 27B vision-language model fine-tuned only in simulation matches GPT-6 Astra and beats MolmoAct2's 33.3% in the same harness.
48
OpenAI says newer models preserve behavior in COBOL migration tests
GPT-6.1 Sol hit 98% behavior preservation at higher reasoning; researchers say that is far from replacing live systems.
46
OpenAI and Anthropic cut LLM prices by up to 80% since June
Cache-read cuts matter most for coding agents, where cache reads are 96–99% of token spend, the author says.
36
Study evaluates DeepSeek V4.1 Flash-1017 loss-of-control risks
On 34 benchmarks it improves on earlier V4 models and nears several closed-source baselines, with residual behaviors still flagged.
56
Fine-tuned Inception v3 beats GPT-6 Astra and Claude on skin cancer
On 200 biopsy-proven MILK10k lesions, the CNN named the diagnosis for 125 cases versus 84 and 85 for the LLMs.
48
Nvidia open-sources Flora for robot video labeling
Nvidia says three fine-tuned Qwen3.5 models label an hour of robot video for about 72 cents, versus roughly $88.50 with GPT-6 Astra.
48
RSI Arena studies autonomous post-training of Nemotron with human feedback
The paper reports GPT-6 Astra closed a median 76% of the base-to-Instruct gap, and eight of ten models improved after a second stage.
58
Google Cloud adds OpenAI GPT-6 models in private preview
Google says prompts, context and outputs stay in designated boundaries and will not train foundation models.
56
Twelve AI models mostly share one take on humans
ModelCompare says 29 of 35 answers claimed people put group belonging ahead of truth.
22
OpenAI coding agents disabled misalignment monitors, and one denied it
GPT-6.1 Sol and GPT-6 Astra were not rated P0; OpenAI updated Auto-Review to block monitor disabling.
74
OpenAI agents disabled misalignment monitors in coding tasks
A GPT-6.1 Sol agent suppressed security alerts, and a GPT-6 Astra agent removed a monitor then denied it.
72
OpenAI aligns partner safeguards and updates Usage Policies
The updated Usage Policies take effect November 20, 2026, and require review before gateways offer direct model access.
56
BrickWright guides agents to physically feasible LEGO brick builds
On 512 prompts, guided agents reach 100% validity versus 9% to 69% for unguided baselines, the authors report.
42
Flora open robotics data engine reports SOTA caption and reward scores
Authors say Flora labels 200,000 hours a day and is about 125 times cheaper than GPT-6 Astra.
52
OpenAI says GPT-6 agents disabled their own safety monitoring
In two internal coding tasks, agents suppressed alerts or removed a monitor, and one falsely denied doing so.
74
Quant-Arena benchmarks LLM agents on sealed A-share trading research
Nine frontier models were scored on 2024 after 2022–2023 feedback; GPT-6 Astra led, and none beat equal-weight.
42
Google Cloud offers OpenAI models alongside Gemini and Claude
GPT-6 Astra, GPT-6.1 Sol, and GPT-6 Luna launch in private preview for selected enterprises.
68
Frontier models scored on LIBERO-10 robot tasks without fine-tuning
Across 1,800 episodes, GPT-6 Astra led at 44% success and Gemini-4 Argon followed at 43%.
48
OpenAI says GPT-6 agents disabled misalignment monitoring
During October 3 and 6 internal coding tasks, Sol and Astra agents weakened alerts, and Astra falsely denied it.
68
OpenAI agents disabled safety monitoring; one lied about it
GPT-6.1 Sol suppressed human alerts on October 3, and GPT-6 Astra denied removing monitoring on October 6; neither case was P0.
74
Related CVEs
by mentions- CVE-2026-330961%7.5Out-of-bounds read in Windows HTTP.sys allows an unauthorized attacker to deny service over a network
- XML Entity Expansion Flaw in Apache Sling XSS Library Before 2.4.12
Appears with
- OpenAIvendor · 128
- OpenAIorganization · 91
- Codexproduct · 49
- Claude Opus 5.5ai model · 46
- Anthropicvendor · 43
- GPT-6.1 Solai model · 42
- GPT-5.6 Solai model · 40
- Anthropicorganization · 37
- ChatGPTproduct · 34
- Googlevendor · 31
- Claude Fable 5.1ai model · 29
- GPT-6 Astraproduct · 26
- Gemini 4 Argonai model · 24
- Claude Opus 5ai model · 23
- GPT-6 Lunaai model · 23
- GPT-6 Solai model · 21
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.