This report is from this week’s The Tech Download newsletter. See what you see? You can subscribe here.
2026 has been a turbulent year. Google AI has hopes and dreams.
Things started off well. With the launch of Gemini 3 in late 2025, Google became significantly involved in the frontier AI conversation.
But for most of the rest of this year, it was Anthropic and OpenAI that were at the forefront of AI advancements.
Google will no doubt be hoping that things change with the release of its new flagship model Gemini 4 Argon.
Samuel Boivin | Null Photo | Getty Images
frontier
When Demis Hassabis stepped down as CEO of Google’s AI lab DeepMind in August, I wrote that his new boss, Koray Kavukcuoglu, was taking over the race to catch up with Sam Altman and Dario Amodei’s company.
Google is currently touting this model as outperforming top OpenAI and Anthropic products on several benchmarks.
In the overall benchmark score, Artificial Analysis Intelligence Index, the Gemini 4 trails only the Claude Opus 5.5 and Claude Sonet 5.5 on the leaderboard.
“Benchmarks show that Google Gemini 4 Argon exhibits advanced reasoning for critical tasks, particularly legal reasoning, finance, and other aspects of enterprise knowledge work, including long-running tasks,” Tim Law, director of AI research at IDC, told me.
Lian Jye Su, principal analyst at Omdia, said that while Google was “a laggard in the cybersecurity and coding space,” Gemini 4 brought the company to the forefront in AI, particularly in the cybersecurity space.
cyber security
Google has been actively talking about this.
The company plans to launch new models in phases, starting with trusted cybersecurity partners, while working with the U.S. government to conduct pre-launch safety assessments.
Tulsee Doshi, Google’s Gemini model product lead, told CNBC: “Starting the rollout this way not only gives us more confidence, but also allows us to get trained, cyber-hardened models into the hands of defenders as quickly as possible.”
A number of security incidents caused by rogue AI systems created by OpenAI and Anthropic have created an opportunity for Google to “stake its claim as a trusted model provider for safe and secure AI operations,” Su said.
But Nick Payens, head of AI at Futurum Group, said that while Argon has made Google competitive again in AI, the model doesn’t make it a “leader.”
“But at the very least, Gemini 4 Argon, in combination with other leading frontier models, opens up a new competitive frontier for enterprise knowledge work,” Lo said.
Google said in a release that the new model is already being used internally to optimize memory in Google’s data centers and can free up hundreds of terabytes of memory without purchasing additional hardware. Quantum computing researchers also use this model.
But the real test, as is always the case with new models, is when it sees the light of day and becomes more widely available. The company hasn’t said when that will happen.
“While Google is touting the success of internal use cases, the final proof will be in enterprise production environments once the model is fully released,” Lo said.
News editing
OpenAI has canceled the release of its next model after determining that it does not adequately meet the company’s safety standards.
space x Launched Google AI chips into orbit to power space-based data centers.
In July, OpenAI was sued by a nonprofit group over a cyber attack on its model startup, Hugging Face.
California Governor Gavin Newsom has banned AI “robo bosses” in a landmark state law, overriding an earlier veto.
Former LinkedIn chief executive Rozlansky plans to retire microsoftfollowing the resignation of other executives.
One more thing
Boats carrying tourists on Lake Bled.
Sopa Images | Light Rocket | Getty Images
President Donald Trump’s rebranding of AI as SI (super intelligence for the uninitiated) has had unexpected (though perhaps predictable) consequences.
This led to an “unprecedented” increase in the number of domain (you guessed it, .si) registrations in the small European country of Slovenia.
A spokesperson for Register.si, Slovenia’s official national registration agency, told the BBC that 44,000 .si web addresses were created in September, up from fewer than 2,000 in August.
This is not the first time that AI has caused a surge in the territory of a small country.
Anguilla, a Caribbean island with a population of less than 20,000 people, has experienced a similar boom in website registration in recent years.
Its domain: .ai.
