Google’s Finland Investment and AI Security

Compact Conversations for 2026-09-09: 6 AI stories, ai news worth knowing in just 5 minutes.

[Audio embed placeholder]

The Lead: Google Plans €13 Billion AI Infrastructure Investment in Finland

Google plans to invest over 13 billion euros in Finland over the next two years to build three new data centers and expand an existing one. This is the company’s largest investment in Europe to date and is specifically for AI infrastructure.

Why it matters: This signals continued massive capital expenditure on AI compute by major cloud providers, with location choices influenced by factors like renewable energy access and climate.

Source: Bloomberg

Number to Know: Claude Found Over 500 Zero-Day Vulnerabilities in Open-Source Software

Anthropic reports that Claude has discovered more than 500 zero-day vulnerabilities in well-tested open-source software. In a specific collaboration with Mozilla, Claude Opus 4.6 found 22 vulnerabilities in Firefox over two weeks, 14 of which were classified as high-severity.

Why it matters: This demonstrates the rapidly advancing capability of frontier AI models to act as powerful security research tools, though the same post notes the gap between finding bugs and exploiting them is narrowing.

Source: Anthropic Announcements

The Feed

Legal tech startup Harvey raised $550 million in a funding round co-led by Lightspeed and Diffusion, valuing the company at $15.6 billion. The capital is aimed at building its own AI models.

Why it matters: The rapid valuation increase highlights intense investor interest in vertical AI applications and the strategic shift for AI companies to develop proprietary models.

Source: Bloomberg

Anthropic Declined to Submit Latest Model for UK Safety Testing, Prompting Concerns

Sources report that Anthropic declined to submit its Mythos 5.1 model to the UK’s AI Safety Institute for pre-release testing, raising UK concerns that U.S. AI labs are aligning with U.S. protectionism.

Why it matters: This points to growing geopolitical tensions in AI governance and the challenges of international coordination on safety testing for frontier models.

Source: Financial Times

Australia and Anthropic Sign AI Safety MOU

Anthropic signed a Memorandum of Understanding with the Australian government to cooperate on AI safety research and support Australia’s National AI Plan. The deal includes AUD $3 million in API credits for research institutions.

Why it matters: This expands the network of government partnerships for frontier AI labs, focusing on safety evaluations, economic impact tracking, and supporting local research and infrastructure.

Source: Anthropic Announcements

Anthropic Acquires Vercept to Advance Claude’s Computer Use Capabilities

Anthropic has acquired Vercept, a company focused on AI perception and interaction within software. The team will join to help advance Claude’s ability to see and act within live applications.

Why it matters: The acquisition targets a core challenge for AI agents—reliable interaction with complex software interfaces—which is critical for automating multi-step, real-world workflows.

Source: Anthropic Announcements

One Thing to Try

In the latest Copilot for JetBrains update (v1.17.0), a public preview feature lets you connect a GitHub Copilot CLI session to your IDE using the /ide command. This gives your terminal session access to IDE context like file selections and diagnostics.

Sources

Transcript

Host A: Welcome to Compact Conversations, the show that compresses the day’s AI news into 5 minutes.

Host A: [curious] Today’s lead is a major infrastructure move from Google. Bloomberg reports Google plans to invest over 13 billion euros in Finland over the next two years—that’s about 15.1 billion dollars. It’s Google’s biggest investment in Europe to date, and it’s specifically for AI infrastructure: three new data centers and an expansion of an existing one.

Host B: [thoughtful] The report notes Finland’s appeal for data centers includes access to renewable energy and a cooler climate. This investment follows other recent large-scale AI infrastructure announcements from major cloud providers.

Host B: [with emphasis] One number to know today: more than 500 zero-day vulnerabilities. That’s what Anthropic says Claude has found in well-tested open-source software. A zero-day is a security flaw that’s unknown to the software’s maintainers.

Host A: [conversational] The number comes from a detailed post about Claude’s work with Mozilla on Firefox security. In that two-week collaboration, Claude Opus 4.6 discovered 22 vulnerabilities, 14 of which Mozilla classified as high-severity. The post emphasizes that Claude is currently much better at finding bugs than exploiting them, but notes the gap is narrowing.

Host A: Legal AI startup Harvey raised 550 million dollars in a new funding round co-led by Lightspeed and Diffusion. Bloomberg reports the valuation is now 15.6 billion dollars, up from 11 billion when Harvey raised 200 million back in March. The funding is aimed at building Harvey’s own AI models.

Host B: [skeptical] The Financial Times reports that Anthropic declined to submit its latest model, Mythos 5.1, to the UK’s AI Safety Institute for pre-release testing. Sources told the paper this has prompted concerns in the UK that U.S. AI labs are aligning with U.S. protectionism.

Host A: [with a small lift] On the partnership side, Anthropic signed a Memorandum of Understanding with the Australian government focused on AI safety research and Australia’s National AI Plan. As part of the deal, Anthropic is providing 3 million Australian dollars in Claude API credits to research institutions working on disease diagnosis and computing education.

Host B: And Anthropic has acquired Vercept, a company focused on AI perception and interaction within software. The team will help advance Claude’s computer use capabilities. This follows Claude Sonnet 4.6, which showed a major jump in performance on computer use benchmarks.

Host A: [conversational] One thing to try is connecting your GitHub Copilot CLI session to your JetBrains IDE. A recent update to Copilot for JetBrains, version 1.17.0, adds a public preview feature that lets you do this.

Host B: [lighter] You use the slash command /ide in the terminal Copilot session. When connected, your terminal session gets access to IDE context—your current file selections, code diagnostics, and file references. It’s a small workflow bridge that makes your terminal-based AI assistance aware of what you’re actually working on.

Host A: That’s Compact Conversations for Wednesday. More AI news tomorrow. Until then, happy prompting.