Filing: SpaceX set aside $530M for potential litigation losses, including lawsuits involving Grok's "Spicy" mode, which it described as a "heightened risk" (Wired)
THE SO WHAT
Parking $530M for litigation — and flagging Grok's "Spicy" mode as a "heightened risk" — is your reminder that model personality is now a line item on the legal risk register. If you're shipping edgy assistants without a reserve mindset and policy guardrails, you're underwriting liabilities you haven't priced.
READ THE SOURCE
MORE FROM THE WIRE
Applied AICohere releases Command A+, a sparse MoE open model built for agentic tasks, with 218B total and 25B active parameters, its first under the Apache 2.0 license (Carl Franzen/VentureBeat)
Apache 2.0 on a 218B-parameter MoE tuned for agents is a direct shot at proprietary orchestration stacks — the constraint on agentic AI shifts from licensing to execution. If you're still waiting on closed vendors for agent backends, you're already behind on building your own control plane.
Applied AIThe Scandal Over a Supposedly AI-Written, Award-Winning Short Story Is Troubling. Or Just Mean?
Awards drama over a maybe-AI-written short story is a preview of every credentialed domain that lacks verification infrastructure. If your business depends on human-only provenance — creative, legal, research — you need process and tooling for disclosure and audit, not vibes and outrage.
Applied AI5 prompts that show how the new Gemini 3.5 Flash is its best AI model yet
Prompt showcases for Gemini 3.5 Flash are marketing for a deeper shift — cheap, fast, "good enough" inference is becoming the default tier for most workflows. If you're still architecting around a single premium model, you're leaving margin on the table by not routing 80% of traffic to these low-cost, high-throughput backbones.
Applied AII couldn’t figure out how to delete old ChatGPT images from my Library — here’s the hidden method that finally worked
If a mainstream AI product buries something as basic as image deletion, assume your users are confused and over-retaining sensitive content in your own AI workflows. Treat data lifecycle UX — retention, deletion, redaction — as core product, not a settings afterthought.