London, April 20, 2026, 09:44 BST
On Monday, Anthropic’s Mythos model drew fresh scrutiny from regulators in Australia and South Korea, highlighting the AI system’s growing profile in a day packed with news across banks, chips, data centers and robotics. Australia’s securities regulator said it is “closely monitoring” Mythos alongside its international counterparts. South Korea’s Financial Supervisory Service, for its part, called in financial companies to discuss risks tied to the Mythos AI. Reuters
The latest wave of AI tools isn’t stopping at spitting out text or code. They’re edging closer to “agentic AI” — tech that starts planning and acting on its own, with minimal nudging from people. That shift is forcing banks and regulators to wrestle with a tougher dilemma: could the same AI that flags vulnerabilities also help hackers exploit them? Reuters
Anthropic pitches Mythos as a defensive effort. With Project Glasswing, certain partners get access to Claude Mythos Preview, aiming to hunt down and patch vulnerabilities in critical software. Launch partners, the company said, include key names from the cloud, chip, and cybersecurity sectors. Anthropic has also promised up to $100 million in usage credits and $4 million in donations for open-source security groups.
Chip stocks followed suit. Morgan Stanley noted agentic AI has the potential to redirect more investment toward central processing units (CPUs) and memory, despite ongoing robust demand for graphics processing units (GPUs)—the chips currently leading AI training and inference. “As AI transitions from generation to autonomous action, the computing bottleneck is shifting towards CPU and memory,” the firm said. Reuters
Google is negotiating with Marvell Technology on a pair of new AI chips, according to The Information—a move that signals Alphabet is pushing for tighter grip on its model-running hardware. The plans include a memory processing unit meant to complement Google’s own tensor processing unit, or TPU, and another chip: a fresh TPU built for powering AI models. Reuters was unable to confirm the report right away, noting that neither Google nor Marvell had responded to requests for comment.
Capital keeps flowing into infrastructure. NEXTDC out of Australia announced plans to raise A$1.5 billion ($1.07 billion) to accelerate its S4 Sydney data centre buildout, after seeing contracted utilisation jump roughly 60% to 667 megawatts. CEO Craig Scroggie described the move as a “unique opportunity”—expanding contracted capacity and reducing risk on Western Sydney projects ahead of any potential private-capital deals. Reuters
The chip supply chain’s momentum keeps building. Renishaw, the British engineering group, bumped up its 2026 revenue and profit outlook, crediting a surge from semiconductor and electronics manufacturing equipment clients. Reuters pointed to demand coming out of AI data centres and autonomous vehicles. Renishaw now projects adjusted pretax profit between 145 million pounds and 165 million pounds, topping its earlier range of 132 million to 157 million pounds.
Europe’s regulatory debate intensified. German Chancellor Friedrich Merz called for looser rules on industrial AI compared to consumer applications under the EU framework, saying he’d work to reduce what he described as red tape, and, where feasible, carve out industrial AI from what he termed a “regulatory straightjacket.” Germany is looking to close the AI gap with the United States and China, aiming to use the technology to boost productivity and cut expenses. Reuters
The China story played out on the streets: dozens of humanoid robots built by Chinese firms lined up for the Beijing half-marathon. An Honor-developed robot clinched the win, clocking 50 minutes and 26 seconds. Reuters noted the field swelled to over 100 teams this year—just 20 competed last year—and close to half the entrants ran on autonomous navigation instead of remote control.
Still, the day’s headlines flagged clear pressure points. Regulators could pull back or restrict access to advanced models if threats to the financial system or cybersecurity remain unresolved. Chip supply negotiations stall easily, sometimes before a single wafer rolls off the line. And the data-center business? It’s hostage to energy prices, hardware costs, customer demand, and a patchwork of local planning codes. Siemens CEO Roland Busch, as reported by Bloomberg, put it bluntly: unless the EU updates its approach, the company will steer its AI investment toward the U.S. and China. Regulation, in other words, is fast turning into a competitive factor—no longer just a compliance headache.
Today’s AI tech headlines weren’t just about rolling out a new product. The story now is control—access to frontier models, chip manufacturing, who’s putting up the cash for data centres, and who’s shaping the regulatory landscape. That’s where the real competition in artificial intelligence is playing out next.
