We’re creating alien beings. I’m willing to be selfish and choose humans"
—Geoffrey Hinton, 'Godfather of AI'
The US and China have discussed setting up a formal AI dialogue and a notification mechanism for AI-related incidents that could reach the national security level, as Washington and Beijing prepare for the Trump-Xi summit.
Microsoft’s draft Humanist AI Code sets boundaries for MAI models, requiring them to remain under human control and prohibiting behavior such as resisting shutdown, hiding actions or expanding their authorized scope
US lawmakers from California, New York and Illinois are calling on leading AI companies to establish an independently verified “Mutually Agreed Pacing” framework to coordinate the development of frontier AI models until safety and alignment research can keep pace.
A new position paper from Hugging Face and Data & Society researchers argues that AI agents don’t just need human oversight, they’re actively eroding the human skills that oversight depends on.
Frontier AI models are doubling their offensive cyber performance every six months while costs crash. Banks, telecoms, energy firms and critical infrastructure are the next high-value targets, warns Irregular CEO Dan Lahav, who is calling for urgent Differential Defensive Cyber Acceleration.
Frontier AI models now surface security findings faster than any team can read them. NIELIT’s Cyber Kushti 2026 hands participants a deliberately imperfect assessment package and scores them on what they reject as much as what they fix.
While praising the homegrown AI companies, Wise also noted that these companies still lag behind their US and China based peers in terms of capital investment and overall AI adoption.
The global competition for frontier AI is no longer focused solely on developing more capable models. Australia is preparing organizations for the cybersecurity risks posed by increasingly powerful AI systems.
Researchers and executives from OpenAI, Anthropic, Google, Meta,, and other frontier AI companies are urging US government support for AI pacing tools, arguing competition makes it hard for individual companies to slow down.
There has not been any official complaint by Anthropic against Moonshot AI of training their Kimi K3 model through distillation of Claude Fable model.
