Anthropic disables top-tier AI models after US order limiting foreign access ↗
Anthropic disabled Fable 5 and Mythos 5 for every customer after a US export-control order required foreign nationals to lose access. All its other models remain available.
Officials reportedly raised concerns about a narrow jailbreak that could help uncover software vulnerabilities. Anthropic disputes the evidence and says the response is wildly disproportionate - a remarkably blunt assessment.
OpenAI faces investigation from state attorneys general ↗
A coalition of state attorneys general has opened an investigation into OpenAI. A New York subpoena seeks information about advertising, user retention, model sycophancy and the handling of consumer and health data.
The inquiry also covers OpenAI’s treatment of minors and seniors. OpenAI says it will cooperate, pointing to parental tools, age prediction and additional protections for vulnerable users.
Meta reportedly moves to unwind $2B Manus deal after Beijing’s demand ↗
Meta has reportedly begun dismantling its Manus acquisition, halting data sharing and cutting the agentic AI startup off from internal systems. The separation follows a divestiture order from Beijing on national-security grounds.
Manus’ founders have discussed raising roughly $1 billion to reclaim the company, potentially through a Chinese joint venture and an eventual Hong Kong listing. A landmark AI deal is becoming geopolitical spaghetti... or so it seems.
KPMG pulls report on AI usage due to apparent hallucinations ↗
KPMG removed a report about agentic AI after several organisations said its claims about their technology use were false or misleading. GPTZero traced numerous inaccuracies to apparent AI hallucinations.
UBS, the NHS, Swiss Federal Railways and Transport for London challenged material attributed to them. KPMG has launched an investigation and stressed that AI-generated work should receive human verification - the irony is almost skyscraper-sized.
The future of Hollywood isn’t feeding prompts into vanilla gen AI models ↗
Tribeca’s AI-assisted films revealed a sharp divide between incoherent generated montages and projects in which artists used models as controlled production tools. So, less magic prompt box... more specialist paintbrush.
Dear Upstairs Neighbors used original concept art to train custom versions of Google’s Veo and Imagen models, preserving a deliberate visual style. The stronger projects suggested that human direction still matters enormously - it turns out the machine needs a filmmaker.
Gore Verbinski says we need an AI “rating system” for movies ↗
Director Gore Verbinski called for a rating system that reveals how extensively films use AI, particularly when scripts or other major creative elements are machine-generated.
He is not entirely opposed to the technology and acknowledged that independent filmmakers may use it to overcome budget limits. His larger concern is transparency - audiences should know what is human, synthetic or suspended somewhere uneasily in between.
FAQ
Why did Anthropic disable its most advanced AI models?
Anthropic disabled Fable 5 and Mythos 5 after a US export-control order required foreign nationals to lose access. Officials reportedly tied the restriction to concerns about a narrow jailbreak that could aid software-vulnerability discovery. Anthropic disputes the evidence behind the decision and argues that disabling the models for every customer is a disproportionate response.
What is the OpenAI investigation by state attorneys general examining?
The investigation covers OpenAI’s advertising, user-retention practices, model sycophancy and handling of consumer and health data. It also examines protections for minors, seniors and other potentially vulnerable users. OpenAI says it will cooperate and has pointed to measures including parental tools, age prediction and additional safeguards.
Why is Meta reportedly unwinding its Manus acquisition?
Meta reportedly began separating from Manus after Beijing ordered divestiture on national-security grounds. The company has halted data sharing and withdrawn the agentic AI startup’s access to internal systems. Manus’ founders have discussed raising about $1 billion to reclaim the business, possibly through a Chinese joint venture and a future Hong Kong listing.
What does the KPMG AI hallucination controversy show?
The incident underscores the risks of publishing AI-assisted research without rigorous human verification. Several organisations challenged claims about their use of technology, while GPTZero identified numerous apparent hallucinations. KPMG removed the report and opened an investigation, reinforcing the need to check generated material directly against reliable sources before publication.
How are filmmakers using generative AI more effectively?
The stronger Tribeca projects treated generative AI as a controlled production tool rather than a one-prompt substitute for filmmaking. Dear Upstairs Neighbors used original concept art to train customised versions of Google’s Veo and Imagen models. This approach helped preserve a deliberate visual style while keeping human creative direction at the centre of the production process.
Why are filmmakers calling for an AI rating system for movies?
Director Gore Verbinski argues that audiences should know how extensively a film uses AI, particularly when scripts or other major creative elements are machine-generated. He is not entirely opposed to the technology and recognises its value for independent filmmakers working with limited budgets. His proposal centres on transparency about what was created by people and what was generated.