News Archive
Pick a date to read that day's stories.
Days with a blue dot have stories.
- 01 of 10ANTHROPIC
Alibaba used 25,000 fake accounts to copy Claude
Alibaba-linked groups made 25,000 fake accounts. They sent Claude millions of questions to copy its knowledge. This is the biggest known AI copying case.
OUR TAKEAlibaba's approach to intellectual property law appears to be 'if in doubt, just ask the thing 28 million times'.
Read more → - 02 of 10OPENAI
White House delays GPT-5.6 release for security checks
The White House asked OpenAI to delay its GPT-5.6 release. This allows the government to perform security checks on the new model. A public launch is now expected in July.
OUR TAKEThe most powerful AI model in the world is currently gatekept by the same government that once banned TikTok dances.
Read more → - 03 of 10GOOGLE
Four top Google AI scientists quit in one week
Four top AI scientists left Google in one week. Their departure caused Google's parent company stock to drop significantly.
OUR TAKEGoogle paid $2.7bn to lure one back two years ago, in hindsight, a very expensive taxi.
Read more → - 04 of 10META
Meta is the only US AI firm refusing government safety checks
Meta is currently the only major US firm declining government safety reviews for its powerful AI models. Other big companies are allowing these voluntary checks. This raises concerns about how Meta's models might be used.
OUR TAKEMeta's line is 'we respect your process, we just won't use it', bold, given the audience.
Read more → - 05 of 10RESCUE
AI drone finds lost hikers in Australian national park
A special drone helped rescue two lost hikers in Australia. It used heat sensors at night to find them quickly.
OUR TAKEReassuring that, while reshaping capitalism, AI still occasionally finds a lost hiker in the dark.
Read more → - 06 of 10NVIDIA
Amazon AI system trains itself and fixes its own broken test
Amazon's AI trained itself, even fixing its own faulty test. It improved without human help, a rare example of AI self-correction.
OUR TAKEAn AI fixing its own evaluation flaws unsupervised is either progress or a film's opening scene.
Read more → - 07 of 10US GOV
US bill would force AI firms to report safety incidents by law
A new bill requires advanced AI companies to report serious safety issues. Currently, no law makes them share such problems. This bill helps the government see these incidents.
OUR TAKEAmerica's answer to opaque AI safety failures is a bill lobbyists will spend years diluting.
Read more → - 08 of 10CHINA
Moonshot AI releases powerful open coding model Kimi K2.7
Moonshot AI released a strong new coding tool. It helps programmers write and check software, performing better than earlier versions. This free tool is now available to everyone.
OUR TAKEChina's open-source coding models now ship faster than Western firms can schedule the response meeting.
Read more → - 09 of 10LEGACY
OpenAI quietly retires GPT-4.5 from ChatGPT today
OpenAI replaced its older AI model in ChatGPT with a newer version. This change means the public now uses the latest technology.
OUR TAKEA quiet Saturday farewell to the model that made everyone's uncle ask about AI at Christmas.
Read more → - 10 of 10POWER
Oracle sues Wisconsin over strict data centre finance rules
Oracle is suing Wisconsin. The state's new rules for computer centers could cost the company an extra $100 million. Oracle wants the court to stop these rules.
OUR TAKEOracle is suing a state for asking financial assurances on a $15bn build, power balance summarised.
Read more →