news
The latest in news from spAIsee.
Can political alignment be edited out of an open AI model?
Hirundo says it made Alibaba’s Qwen more politically neutral without reducing capability, but independent testing must determine whether the change holds across unfamiliar prompts and real-world evaluations.
Beam puts inference efficiency at the center of open-model competition
Reflection AI’s Beam targets the open-model efficiency race, promising frontier-style reasoning and coding performance with far lower inference costs, though its benchmark and compute claims remain unverified.
Claude Haiku 5.5 brings frontier benchmarks into the low-cost AI race
Anthropic’s Claude Haiku 5.5 enters the low-cost AI race with a claimed 90% price reduction, challenging OpenAI’s GPT-6 Luna for high-volume enterprise workloads and everyday automation.
Microsoft wants Windows to become an AI workstation
Microsoft is turning Windows into an AI workstation, combining local models, cloud services and controlled agents while launching powerful Surface hardware for developers and addressing the security risks of autonomous software.
Tony Fadell says AI gadgets failed before they earned users’ trust
Tony Fadell explains why Rabbit R1, Humane AI Pin and Limitless Pendant struggled, arguing that AI gadgets must solve clear problems and earn trust before requesting access to users’ private lives.
OpenAI’s GPT-6 rollout puts product execution at the center of AI competition
OpenAI’s GPT-6 rollout pairs a new model with Intelligent UI, testing ChatGPT’s product execution, subscription strategy and ability to turn advanced AI into everyday workflows.
Google Opens a Public Tool to Detect AI-Generated Images, Video and Audio
Google has launched SynthID Detector, a public tool that checks images, video and audio for invisible AI watermarks while warning users that negative results do not prove authenticity.
Tab Wants to Turn Messages Into a Personal AI Command Center
Tab is emerging from stealth with a Messages-based AI assistant that can browse, call businesses, use connected accounts, create files and pause for approval before purchases.
A New Human Authorship Mark Tests the Value of Trust in Books
Books By People is launching an “organic literature” certification mark to signal human authorship as AI-generated prose spreads, raising questions about proof, software and trust.
TwelveLabs Bets on First-Person Video to Close Robotics’ Data Gap
TwelveLabs’ Pegasus 1.6 analyzes first-person work videos to extract actions, objects, hand movements and task boundaries, targeting robotics’ costly shortage of structured training data for physical AI.
Claude Moves Into Google Workspace, Putting Gemini’s Home Advantage to the Test
Anthropic is bringing Claude into Google Docs, Sheets and Slides, challenging Gemini’s built-in Workspace advantage and turning AI competition into a fight over everyday office workflows, permissions and user habits.
When AI Safety Audits Fail Before the Agent Does
AI safety audits can fail before agents do when they search incomplete evidence, trust model summaries, or miss infrastructure signals. This analysis examines Anthropic’s and OpenAI’s incidents and explains how to build harder-to-fool evaluation pipelines.