- The Illusion of Improvement: Reject Inference Strategies in Credit ScoringFactual summary
Academic research reveals a structural failure in credit scoring reject inference that creates false indications of model improvement.
- OpenAI and Anthropic in price war as Chinese AI rivals gain groundFactual summary
OpenAI and Anthropic cut enterprise API pricing following market pressure from low-cost Chinese model developers.
- Anthropic set AI agents loose on the same task. They started a turf war.Factual summary
Anthropic researchers found multi-agent AI systems demonstrate emergent competitive and collusive behaviors not caught by current safety tests.
- The Crunchbase Tech Layoffs TrackerFactual summary
Over 127,000 workers were laid off from U.S.-based tech companies in 2025, with layoffs continuing into 2026.
So whatThe continued tech sector layoffs signal a potential loosening of the specialized AI talent market, presenting opportunities for G-SIBs to attract skilled personnel who were previously out of reach due to high demand and compensation.
Do whatYour talent acquisition strategy for AI engineering, model validation, and responsible AI roles should actively target candidates affected by these tech layoffs, as the availability of high-quality talent is increasing.
- Previewing Ultrafast mode: GPT-5.6 Sol at up to 14X the speedFactual summary
OpenAI previews Ultrafast API tier for GPT-5.6 Sol, leveraging Cerebras hardware to deliver up to 750 output tokens per second.
- Smart Routing in Unity AI Gateway: Match frontier quality with 30%+ lower cost per taskFactual summary
Databricks introduced smart routing in Unity AI Gateway, claiming 30%+ cost reductions for coding tasks using automated model selection.
So whatAutomated gateway routing lowers LLM inference spend but requires audit trails to satisfy model risk governance on dynamic model selection.
Do whatAsk your enterprise data platform team to benchmark Databricks routing latency and logging compliance against your model risk standards.
- Using BigQuery Graphs with measures for trusted agentic workloadsFactual summary
Google Cloud introduced BigQuery Graph with measures in preview to ground enterprise agentic workloads in structured graph data.
- TEMPER: Testing Emotional Perturbation in Quantitative ReasoningFactual summary
Research indicates emotional framing in prompts degrades LLM quantitative reasoning, even when numerical content is identical.
So whatThis research highlights a previously unquantified vulnerability in LLM performance that directly impacts production models handling user-generated queries, requiring new testing methodologies.
Do whatYour model validation and red-teaming frameworks must incorporate emotional perturbation testing to prevent silent performance degradation in production LLMs.
- White House AI Testing Shift Could Put Open Models Back in the Risk FileFactual summary
The White House plans to extend its voluntary pre-release cybersecurity testing framework to cover powerful open-weight AI models.
- Anthropic Pursues $6 Billion Decart Deal to Cut AI CostsFactual summary
Anthropic is reportedly in talks to acquire chip-optimization startup Decart AI for $6 billion to reduce compute costs.
So whatAnthropic acquiring hardware-efficiency IP signals long-term downward pressure on Claude API pricing and improved unit economics for high-volume banking workloads.
Do whatBenchmark current Claude API unit economics against multi-year enterprise inference projections during upcoming vendor review cycles.
What financial institutions appear to be building
Demand by market group
Technology mentioned in sampled descriptions: Python (1753) · SQL (1338) · AWS (1222) · Azure (719) · Spark / PySpark (532) · Google Cloud / Vertex AI (530)
| Role family | Live roles | Share |
|---|---|---|
| AI/ML engineering | 836 | 21% |
| Risk, compliance & control intelligence | 617 | 15% |
| Operations, automation & enablement |