🛡️ Anthropic Opens Claude Security to All

Anthropic is bringing AI deeper into cybersecurity with Claude Security, now in public beta. The tool helps teams find and fix software flaws faster by reasoning through code, data flows, and file connections more like a human security researcher. Let’s dive in!

In today’s insights:

LATEST DEVELOPMENTS

Evolving AI: Anthropic moved its AI-powered vulnerability scanner into public beta, letting cybersecurity teams hunt and patch software flaws at scale.

Claude Security is part of the Claude Enterprise tier and shifts the model from pattern matching to reasoning. It traces data flows, reads source code, and studies how files interact, much like a human researcher would. Each finding comes with a confidence score, exploitability estimate, and a suggested fix that can be applied directly inside a Claude Code session. The beta also adds scheduled scans, dismissal notes for triage, and CSV and Markdown exports for audit pipelines. The launch follows Anthropic's Project Glasswing coalition and its Mythos model work.

AI is now writing a huge share of enterprise code, and that code is showing up to 30% more vulnerabilities than what humans wrote on their own. Meanwhile attackers are using AI too, and IBM clocked a 44% jump in attacks that started with exploiting public-facing apps last year. So you've got bugs being created faster, found faster, and weaponized faster, while human review teams are stuck moving at human speed. Tools like Claude Security are basically the defense side trying to catch up to its own offense. The real question is whether AI can clean up the mess AI is making.

Last week, Viktor wrote a campaign brief, built a landing page, opened a pull request, generated a board-ready PDF from live Stripe data, and sent a follow-up email to a churned customer. All from Slack. Same colleague that also pulls your reports and monitors your dashboards. 5,700+ teams. 3,000+ integrations.

Start free. $100 in credits →

Evolving AI: Google is rolling out Gemini to vehicles with Google built-in, replacing Google Assistant with a more conversational AI co-pilot.

Google announced the Gemini-for-cars rollout, expanding on GM's reveal a day earlier. Existing compatible vehicles get the upgrade through software updates, not just new models. Drivers can ask layered questions like finding a sit-down restaurant with outdoor seating along their route, then follow up about parking or menu options. A beta feature called Gemini Live, activated by saying "Hey Google, let's talk," supports open-ended conversation. Future updates will deepen ties with Gmail, Calendar, and Google Home.

Cars are becoming one of the largest deployment surfaces for conversational AI outside the phone. Putting Gemini in millions of dashboards turns routine drives into live tests of voice AI at scale, raising fresh questions about driver attention, data collection, and how much of the in-car experience automakers are willing to hand over to a single tech platform.

Speak your prompts into ChatGPT or Claude and get detailed, paste-ready input that actually gives you useful output. Wispr Flow captures what you'd cut when typing. Free on Mac, Windows, and iPhone.

Try Wispr Flow free

Evolving AI: OpenAI launched Advanced Account Security for ChatGPT, partnering with Yubico on co-branded hardware keys to fight phishing.

The new program lets ChatGPT users tie their accounts to a physical security key, blocking attackers even if passwords are stolen. OpenAI is targeting people whose conversations carry real risk, including political dissidents, researchers, and enterprise users storing sensitive data in chat sessions. Reports show cybercriminals are increasingly targeting chatbot accounts for extortion. There is a tradeoff though. If a user loses their key, OpenAI cannot recover the account and past conversations may be gone for good.

People are pouring secrets into ChatGPT every day, from therapy-style chats to corporate strategy, and attackers have noticed. Phished accounts already turn up on dark web markets, and AI-related breaches like the Mercor and Vercel incidents this spring showed how fast one stolen login can cascade into something much bigger. A physical key sitting on your desk is one of the few defenses phishing kits genuinely cannot beat. The bigger shift here is that AI accounts are quietly becoming as sensitive as your bank login, and the industry is starting to treat them that way.

Most people are using AI wrong, and this video shows exactly why. It breaks down how small changes in your prompts can dramatically improve the quality of outputs, from content to code. If you want better results from tools like ChatGPT, this is a quick, practical upgrade.

QUICK HITS

🖥️ AWS launches Amazon Quick desktop AI assistant that works across your applications, tools, and data.

👨‍⚖️ Elon Musk testifies in a case that could change the path of AI.

🧬 Groundbreaking new study shows real-time AI platform better at diagnosing cancer than biopsy.

🗣️ Amazon's 'Hear the highlights' shopping feature now lets you ask questions and get real-time answers.

📱 Enabling privacy-preserving AI training on everyday devices.

 *partner link