Category intelligence
All 41 current items, analyzed and ranked.
Executive synthesis
Primary evidence
By Unknown
Official Google DeepMind blog post introducing Gemini 3.8 Live and 3.8 Live Extended Thinking audio models for developers. The announcement ties the new voice models to top speech-to-speech leaderboard performance and aggressive per-hour pricing.
By Rebecca Bellan
OpenAI confirmed it has spent weeks in AI safety discussions with Anthropic and Google DeepMind, even as the Trump administration dismisses safety concerns and emphasizes pace with China. The talks involve senior leaders including Sam Altman, Dario Amodei, and Demis Hassabis.
By Unknown
An AEF-1 standard for third-party evaluators has emerged, with xAI, OpenAI, and Anthropic cosigning. The proposal, detailed in a personal blog post by Dario Amodei, calls for embedded evaluator teams with employee-like access to verify safety commitments and assess training pipelines.
By Maximilian Schreiner
The Gates Foundation is committing at least $1 billion over two years to broaden AI access in health, education, and agriculture, with Bill Gates warning that more than 90% of training data is English-source and speech recognition fails ~60% of the time in Yoruba. Gates frames markets as a poor guarantor of equal AI opportunity.
By Richard Lawler
An NYT/Siena poll of 1,503 likely voters found 61% oppose construction of AI data centers versus only 14% strongly supporting it. Opposition crosses partisan lines, including 45% opposition among 2024 Trump voters, suggesting data center politics could shape midterm elections.
By Matthias Bastian
Anthropic's decision to store usage logs from its flagship Fable model for 30 days spooked Palantir, Nvidia, and Booz Allen Hamilton into pulling back from sensitive workloads. The piece argues that AI labs' data policies have not yet solved enterprise trust, even when training opt-outs are formally in place.
Enterprise AI demand is shifting toward cheaper, unregulated open-source models from China while US frontier labs push for a development slowdown. The piece highlights the tension between safety-driven deceleration and competitive market pressure.
By Luca Ittimani and Caitlin Cassidy
Australia's Labor government is considering proposals that would give AI companies default access to Australian creative works for training, revealed as senior OpenAI personnel met with ministers. The party faces backlash from creative industry advocates who warn the move undermines copyright protections.
By Julie Bort
An early Anthropic employee and former METR COO launched AIUC (Artificial Intelligence Underwriting Company), which raised a $40 million Series A led by Ribbit Capital to develop tools that constrain rogue AI agents. The startup applies insurance-style risk underwriting to agent behavior.
By Julie Bort
Salesforce released Koa, a reasoning model fine-tuned from Nvidia's open-weight Nemotron and targeted at sales, marketing, and customer-support workflows. TechCrunch frames it as a competitive threat to frontier AI labs despite being a domain-specific enterprise model.
