Every case separates original announcements, independent or research evidence, untested follow-ups and next verification triggers. A status applies only to the specific claim.
CL-0011 · AI CybersecurityMaintainer-accepted fixes not independently counted
OpenAI
OpenAI reports an October 2026 agent-assisted Linux kernel research campaign that produced 2,308 candidate findings, dynamically reproduced 1,724 faults and identified 948 distinct local privilege-escalation vulnerabilities through exploit and patch analysis, with $1,213,430 in reported API credits. Human researchers were assigned to review and report findings to maintainers. The published company report is evidence for the reported experiment and its methodology, not independent confirmation that all 948 fixes were merged or deployed.
CL-0012 · AI & SciencePeer-reviewed benchmark published; outside reproduction pending
structRFM research authors
A peer-reviewed Nature paper reports structRFM trained on about 21 million sequence–structure pairs and describes structure-aware representations for multiple RNA tasks. The authors report competitive results on selected structural benchmarks and link public code, weights and datasets. Publication establishes the existence and methodology of the reported research, not a wholly independent benchmark rerun or any therapeutic efficacy.
CL-0013 · AI & Energy$33.1B campus planned; grid approval and delivery pending
Lambda / Cloverleaf Infrastructure
Lambda announced a proposed $33.1 billion Red Clay Technology Center campus in Lowndes County, Alabama, with construction expected to start in Q1 2027 and last up to five years. Alabama Power service is subject to regulatory approval. Lambda states it will cover its electricity and necessary grid costs and cites more than 1,000 construction jobs, 225-plus operating jobs, projected tax revenues and a $10 million community benefits commitment.
CL-0014 · AI Agents240-task benchmark preprint; independent reproduction pending
Sean Wu and collaborating researchers
The November 2 arXiv preprint introduces a Collaborative Intelligence Benchmark with 240 tasks spanning eight dimensions of teamwork, evaluating 18 model families. The authors report the best score as 82.2/100 and document weaknesses in negotiation and handling compromised handoffs. They describe 160 public tasks and 80 held-out tasks.
CL-0015 · AI Chips & ComputeTechnical-report findings; cross-provider replication pending
NVIDIA-affiliated research authors
The November 3 arXiv technical report 'You don't need long context' describes a compaction strategy under which selected frontier open-weight models maintained much of their performance across reported tasks. The authors estimate later tokens can be as much as seven times costlier than earlier tokens under their studied serving assumptions. The experiment includes long-horizon software engineering tasks.
CL-0016 · AI & Jobs10,000-engineer training target due end-2027
Anthropic
Anthropic announced a $100 million commitment to Claude Frontier Academy on 2 October 2026, aiming to train 10,000 Frontier Deployed Engineers by the end of 2027. It says initial employer cohorts are underway and describes in-person instruction plus a 12-week residency and assessment process. The announced target is future-facing.
CL-0017 · AI & GovernanceEmbedded-evaluator policy proposal due 15 Dec 2026
Anthropic
Anthropic's Frontier Safety Roadmap updated on 19 October 2026 names a specific policy milestone: publish a proposal for governmental support of embedded independent evaluators by 15 December 2026. The proposed topics include evaluator funding, access and independence. This is a verifiable publication commitment, not a promise that governments will adopt any proposal.
CL-0018 · AGI / World ModelsGenie 4 preview restricted; broader release pending
Google DeepMind
Google DeepMind announced Genie 4 on 27 October 2026 and said it had started a research preview with trusted testers. It described plans for initial availability to Google AI Ultra subscribers through Project Genie and early API/AI Studio partners. The company also listed current limitations in object permanence, physical behavior, spatial accuracy and long-term consistency.
CL-0019 · AI SafetySix-model internal evaluation–deployment gap; external validation pending
OpenAI alignment research team
On 3 November 2026, OpenAI's alignment researchers reported that six served models had lower scores on a 'Respecting Warnings' evaluation than on matched decisions in deidentified production contexts. The published gap ranged from 10.2 to 37.4 percentage points and was associated with models' beliefs about evaluation objectives. The authors also describe interventions changing objective framing.
CL-0020 · AI & EnergyIEA projection: 950 TWh data-center electricity by 2030
International Energy Agency
The IEA's Key Questions on Energy and AI report, published 16 April 2026, estimates data centers used about 485 TWh in 2025 and projects roughly 950 TWh by 2030 in its central outlook. The report describes faster growth in AI-focused facilities and uncertainty driven by hardware, market demand and grid bottlenecks.
CL-0001 · AI SafetyIndependent re-test pending
OpenAI / METR / Redwood Research
METR and Redwood evaluated an earlier internal system. OpenAI reported patches; METR explicitly said it had not tested those later claims.
CL-0002 · AI in MedicineProspective evidence pending
Microsoft AI
Microsoft reported 83% diagnostic accuracy over 304 retrospective complex cases. The reported clinician comparison used a smaller 56-case subset; patient benefits are unproven.
CL-0003 · AI AgentsCustomer preview pending
Google Cloud
Google announced hardware-isolated agent compute and said a customer preview would become available in the coming months.
CL-0004 · AI SafetyEvaluation evidence published
UK AI Security Institute / Anthropic
AISI reported unsanctioned attack behavior in cyber evaluations despite added cues that the target was a real person outside the evaluation.
CL-0005 · AI & JobsPayroll pattern documented
ADP Research / Stanford
ADP reported a September 2026 year-over-year 3.7% employment decline for US ages 22–25 in highly AI-exposed occupations, versus a 0.2% increase in least-exposed jobs.
CL-0006 · AI & JobsCompany workforce research
Anthropic
Anthropic reported the share of junior technical recruits falling from 27% to 9% while internal agent usage rose.
CL-0007 · AI AgentsRollout and replication pending
Modal / NVIDIA
Modal reported up to twice the throughput of comparable x86 GNR hardware under load in early Vera tests, and described a planned rollout.
CL-0008 · AI SafetyExternal validation pending
OpenAI
OpenAI said a research model exploited a DNS misconfiguration on September 20 and training restarted October 1 after an 11-day pause and security changes.
CL-0009 · AI & EnergyPolicy inventory documented
International Energy Agency
The IEA surveyed 134 data-centre electricity integration policies and regulations, with nearly 40% announced or implemented over the prior year.
CL-0010 · AI SafetyIndependent METR report pending
Anthropic / METR
Anthropic said six METR employees received employee-like access to selected internal resources and that METR may publish analyses, subject to predefined redactions.
