Japan Probes Musk’s Grok AI Over Inappropriate Images

Japan has launched a probe into X and its AI chatbot Grok, becoming the latest country to scrutinise the service over the generation of inappropriate images linked to Elon Musk.

Economic Security Minister Kimi Onoda said the Cabinet Office has asked X to implement immediate improvements but has yet to receive a response. She warned that Japan would consider “all possible options, including legal measures,” if the situation does not improve.

Developer xAI said this week it had updated Grok to restrict editing of images of real people in revealing clothing and to block such image generation in jurisdictions where it is illegal, without naming specific countries.

Japan’s move follows similar actions by the UK and Canada, as regulators worldwide seek to curb Grok’s ability to generate sexualised images of women and minors. Malaysia and Indonesia have already temporarily blocked access to the chatbot over explicit content concerns.

AI Hyperscalers Seen Driving Surge in U.S. Corporate Bond Supply in 2026

U.S. corporate bond issuance is set to rise sharply in 2026, with artificial intelligence hyperscalers expected to be a major driver as they ramp up spending on data centres and computing power, analysts say.

According to a report from Barclays, total U.S. corporate bond issuance could reach $2.46 trillion in 2026, up nearly 12% from $2.2 trillion in 2025. Net issuance is forecast at $945 billion, a 30% increase from last year, with AI-related capital expenditure seen as the biggest upside risk.

Barclays said the rise in supply would be driven less by financial firms and more by non-financial companies, particularly large AI-focused technology groups that may need to fund “jumbo” bond deals to support rapid expansion.

The five major AI hyperscalers — Amazon, Google, Meta Platforms, Microsoft and Oracle — issued about $121 billion in U.S. corporate bonds last year, far above their 2020–2024 average of $28 billion, according to BofA Securities.

BofA expects hyperscaler borrowing to accelerate further, estimating the group could raise around $140 billion a year over the next three years, potentially exceeding $300 billion annually. That would put them on par with the largest U.S. banks as bond issuers and make them some of the biggest names in the investment-grade market.

Hyperscalers already dominated issuance in 2025. According to MUFG, four of the five largest U.S. high-grade bond deals last year came from AI-driven tech firms, including Oracle’s $18 billion sale and Meta’s $30 billion deal — the largest ever non-M&A investment-grade bond offering.

The surge in supply has begun to widen credit spreads, pushing investors to hedge AI-related risks through credit default swaps. MUFG data shows the cost of insuring hyperscaler debt has risen since mid-2025, with Oracle’s five-year CDS more than tripling since its September bond sale.

Trump’s Approval of Nvidia AI Chip Sales to China Sparks Backlash in Washington

U.S. lawmakers and former national security officials are pushing back against President Donald Trump’s decision to allow Nvidia to sell its H200 artificial intelligence chips to China, warning the move could weaken America’s AI advantage and bolster Beijing’s military capabilities.

The Trump administration on Tuesday formally approved China-bound sales of Nvidia’s H200, the company’s second most powerful AI processor, setting the stage for shipments to resume despite long-standing concerns in Washington over national security risks.

At a congressional hearing, Matt Pottinger, a former White House Asia adviser during Trump’s first term, said the administration was on the “wrong track” and argued that selling advanced AI chips to China would “supercharge Beijing’s military modernization.” He warned the chips could enhance capabilities ranging from nuclear weapons development to cyber warfare, autonomous drones and intelligence operations, urging Congress to impose stricter guardrails.

Several Republican lawmakers voiced unease. Congressman Michael McCaul said the United States should not be selling sensitive technology to a country that routinely steals American intellectual property. Democratic lawmakers were more direct, with Congressman Gabe Amo likening the policy shift to “handing our opponents our coordinates in the middle of a battle.”

The administration has defended the decision, with White House AI czar David Sacks arguing that controlled sales could discourage Chinese firms such as Huawei from accelerating efforts to develop rival chips. Pottinger dismissed that logic as unrealistic.

Under the new rules, chips exported to China must undergo third-party testing, and China cannot receive more than 50% of the volume sold to U.S. customers. Nvidia must also certify sufficient domestic supply, while Chinese buyers are required to demonstrate security safeguards and pledge not to use the chips for military purposes.

Some lawmakers acknowledged these safeguards. Congressman Brian Mast, who chairs the House Foreign Affairs Committee, said “know your customer” provisions were significant. Others remained skeptical, noting the rules rely heavily on self-reporting by Chinese buyers and would add strain to the Commerce Department, which oversees export controls.

Adding to the uncertainty, Reuters reported that Chinese customs officials recently indicated Nvidia’s H200 chips were not permitted to enter the country, raising questions about how quickly shipments could actually resume.