AIR Security Raises $50M for AI Firewall

Key Takeaways

• <ul><li>AIR Security raised $50 million to build a firewall for AI agents, with Sequoia and Greenoaks leading the round</li><li>Investors put $435 million into AI agent security startups across 12 deals in five months</li><li>Cloudflare lets websites block AI training crawlers while staying in search results, with Apple, Google, and Microsoft meeting its Accountable certification</li><li>Apple plans to charge for some Apple Intelligence and Siri features in iOS 27, including HomeKit at up to $59.99 per month</li><li>BNY launched an AI platform called Eliza for managing client relationships via AI agents</li><li>ByteDance raised $290 million for Anew Labs, its AI drug discovery unit, valuing it at $1.5 billion while retaining 56 percent ownership</li><li>NVIDIA, Alphabet, and Microsoft each surged over 65% in the 12 months following the July 2023 Fed rate hike</li><li>An attacker spread the Shai-Hulud worm across roughly 100 repositories after hijacking an AI coding session</li><li>Naval Ravikant says AI labs, open source model hosts, and end users should share liability for model behavior</li><li>Lectra launched Apogy, a cloud platform using agentic AI for fashion product development and digital prototyping</li></ul>

AIR Security raises $50M to guard AI agents from risky plugins

AIR Security raised $50 million in a seed round to build a firewall for AI agents. Sequoia and Greenoaks led the funding. The company checks the skills and plugins AI agents use and blocks unsafe ones. It has over 20 customers, mostly in finance and pharma. The team was founded by Unit 8200 veterans Yair Saban and Niv Hoffman.

Investors pour $435M into AI agent security startups

Investors put $435 million into AI agent security companies across 12 deals in five months. AIR Security is one example, building tools that filter unsafe add-ons for AI agents. Other firms like Glow and Zenity tackle different parts of the same problem. Financial services and pharma are buying first due to regulatory pressure. The funding aims to close the gap between fast agent deployment and security readiness.

Naval Ravikant says labs should be liable for AI model behavior

AngelList founder Naval Ravikant says AI labs should be fully responsible for how their models act. He also says hosts of open source models and end users share liability for harm. His approach puts accountability on whoever controls the model at each stage. This differs from the EU AI Act, which uses fixed risk tiers. He believes liability pushes safety work earlier in development.

ByteDance raises $290M for AI drug discovery unit Anew Labs

ByteDance completed a $290 million fundraising for its AI drug discovery unit, Anew Labs. The unit reached a $1.5 billion valuation in its first external fundraising. HSG, IDG Capital, Hillhouse, and 5Y Capital led the round. ByteDance still owns 56 percent of the company after the funding. The spin-off aims to let the unit grow with its own focus and management.

BNY launches AI platform Eliza to modernize finance

BNY CEO Robin Vince introduced an AI platform called Eliza during a talk at Northeastern University. The platform helps employees manage client relationships and meetings using AI agents. It can be used through a phone app. BNY was founded 240 years ago by Alexander Hamilton. Vince said AI development is moving faster than safety alignment and monitoring. He wants capability, alignment, and monitoring to advance together.

Cloudflare launches tool to block AI training while keeping search results

Cloudflare introduced a new setting on September 16 that lets websites block AI training while staying in search results. The feature targets hybrid-purpose crawlers like Applebot, Bingbot, and Googlebot that handle both search indexing and AI training. Cloudflare also launched an Accountable certification, which Apple, Google, and Microsoft have met or committed to. Data shows 17% of websites block training crawlers while fewer than 1% block search crawlers. A granular control for AI summary inclusion ratios is planned for early next year.

3 Resilient AI Stocks That Surged After the Fed Raised Rates in 2023

When the Federal Reserve raised interest rates in July 2023, three AI stocks soared more than 65% in the following 12 months. NVIDIA rose 72%, Alphabet surged 67%, and Microsoft climbed 68%. These companies showed strong resilience during economic uncertainty. They capitalized on growing demand for AI technologies. Past performance does not guarantee future results but these stocks demonstrated impressive strength.

Apple outlines steep AI usage costs that may get worse

Apple Intelligence features in iOS 27 are mostly free for now but Apple plans to charge for some services. HomeKit Secure Video with AI features costs up to $59.99 per month for a five-camera setup. Apple has stated it will implement usage limits on Apple Intelligence and Siri AI features in the future. Increased access to these features will be available for a fee. It is not clear when charging will begin or how much it will cost.

Lectra launches Apogy cloud platform using agentic AI for product development

Lectra launched Apogy, a cloud-based platform that uses agentic AI to reorganize fashion product development. Apogy automates lower-value tasks and helps teams generate concepts from prompts, sketches, or images. It supports 3D simulations and photorealistic renderings for digital prototyping. The platform centralizes product data for secure collaboration among partners and suppliers. Apogy expands Lectra from patternmaking technology to end-to-end product development.

Attacker hijacks AI coding session and spreads Shai-Hulud worm across 100 repositories

Mandiant reported that an attacker hijacked an active AI coding-assistant session at a software-as-a-service provider. The attacker spread the self-spreading Shai-Hulud worm across approximately 100 internal code repositories. The worm stole repository secrets and source code. The attacker also poisoned a package in the company's official namespace causing a second infection. Mandiant recommends checking AI-recommended dependencies against checksums and keeping secrets out of reach of extensions.

Sources

• AIR Security’s $50M Seed Signals a $435M Agent-Security Category Is Forming While Platforms Ship at Full Speed
• AIR Security’s $50M Seed Signals a $435M Agent-Security Category Is Forming While Platforms Ship at Full Speed
• Best Way To Pace Frontier Is To Hold Labs Liable For The Behaviour Of Their Models: Naval Ravikant
• ByteDance completes $290 million fundraising for AI drug unit after its spin-off, sources say
• Another Hamilton is bringing BNY into the AI future
• Cloudflare launches tool letting websites block AI training while staying in search results
• 3 Resilient Artificial Intelligence (AI) Stocks That Did Well the Last Time the Fed Raised Interest Rates
• Apple outlines steep AI usage costs, and it’s probably going to get worse
• Agentic AI Development Solutions
• Attacker Hijacks AI Coding Assistant Session, Spreads Shai-Hulud Across About 100 Repositories