Google to challenge German ruling saying it is liable for AI-generated false claims ↗
Google plans to appeal a Munich court ruling that classifies AI Overviews as the company’s own published content, leaving it legally responsible when those summaries contain false claims.
The case was brought by two German publishers that had been wrongly linked to scams and dubious business practices. Google describes the errors as narrow, but the ruling could place a surprisingly weighty legal backpack on every AI answer engine. (Reuters)
Tech executives to attend G7 summit as leaders address AI, online safety ↗
Leaders from OpenAI, Anthropic, Google DeepMind, Mistral AI, Cohere and Meta are expected to join G7 officials for discussions spanning AI regulation, infrastructure and online safety.
Online protections for children are also on the table. So, yes, the AI industry is taking a seat beside world leaders - though whether that brings genuine accountability or merely an exceptionally ornate lunch remains hazy. (Reuters)
Anthropic Says It’s Taking Claude Fable 5 Offline to Comply With US Government Order ↗
Anthropic disabled Claude Fable 5 and Mythos 5 after receiving a US export-control directive intended to block access by foreign nationals, including Anthropic employees.
The company withdrew access for everyone while contesting the government’s rationale. Officials reportedly cited concerns over a jailbreak, but Anthropic says the demonstrated bypass was narrow and revealed only relatively minor, previously known vulnerabilities... awkward, to put it mildly. (WIRED)
Meta’s new AI unit is reportedly in turmoil ↗
Meta’s Applied AI organisation is facing broad internal frustration, according to employees and company discussions reviewed by WIRED. The roughly 6,500-person group was assembled to support Meta’s superintelligence researchers.
Staff reportedly feel they were funnelled into repetitive model-improvement work amid unclear leadership and erratic priorities. Tensions became rather spectacular when an employee interrupted a large internal presentation with an expletive-laden rant - subtle it was not. (WIRED)
Chinese cybercrime operation that used AI to scam ‘hundreds of thousands of victims’ sued by Google ↗
Google is suing Outsider Enterprise, an alleged cybercrime network accused of using AI-assisted tools to create convincing phishing sites and send scam texts impersonating major brands.
Google says the operation deployed thousands of fake websites, roughly a million fraudulent domains and millions of messages. Its software reportedly allowed people with limited technical skills to generate phishing infrastructure within minutes - cybercrime with the training wheels welded firmly on. (TechCrunch)
FAQ
Why is Google appealing the German AI Overviews ruling?
Google plans to challenge a Munich court decision that treats AI Overviews as content published by Google itself. The ruling followed false summaries that linked two German publishers to scams and questionable business practices. If upheld, the decision could make AI answer providers legally responsible for inaccurate claims generated within their search summaries.
What could the German ruling mean for AI liability?
The decision may establish a broader standard for AI liability by holding a platform responsible for statements produced through its own AI features. Courts in many jurisdictions are still determining whether generated summaries should be treated as hosted third-party content or as direct publication. The appeal could therefore influence how comparable disputes are handled elsewhere.
Why are technology executives attending the G7 summit?
Executives from OpenAI, Anthropic, Google DeepMind, Mistral AI, Cohere and Meta are expected to take part in discussions with G7 leaders. The agenda includes AI regulation, infrastructure, online safety and protections for children. Their attendance gives governments a chance to question major AI developers directly about risks, responsibilities and future policy.
Why did Anthropic take Claude Fable 5 and Mythos 5 offline?
Anthropic disabled both models after receiving a US export-control directive intended to prevent access by certain foreign nationals, including some employees. The company removed access for everyone while challenging the government’s reasoning. Officials reportedly cited concerns about a jailbreak, although Anthropic described the demonstrated bypass as narrow and involving relatively minor vulnerabilities that were already known.
What is causing reported turmoil inside Meta’s AI unit?
Employees reportedly describe unclear leadership, shifting priorities and repetitive model-improvement work within Meta’s Applied AI organisation. The group was created to support the company’s superintelligence researchers and includes roughly 6,500 people. Frustration became especially visible when an employee interrupted a large internal presentation with an angry, expletive-filled complaint.
How are criminals using AI for phishing scams?
According to Google’s lawsuit, the alleged Outsider Enterprise network used AI-assisted software to create convincing phishing websites and scam messages that impersonated major brands. The tools reportedly helped users with limited technical experience build phishing infrastructure quickly. Google says the operation involved thousands of fake sites, around a million fraudulent domains and millions of messages.