NerdNews

Apple has started testing its always-listening AI feature on the Apple Watch Series 12, which includes Live Rewind and Siri Recap. The features raise concerns about privacy and consent, and are currently unavailable in the European Union due to strict tech regulations.

OpenAI has partnered with Cathedral, a cybersecurity startup founded by former DOGE employees, to strengthen its security. Cathedral's team will use OpenAI's models to challenge its defenses and find weaknesses. OpenAI is also increasing its bug bounty rewards to improve security.

CrowdStrike is acquiring Fiddler AI and Dreadnode to expand its AI security capabilities, aiming to capture a share of the burgeoning market for securing AI agents. The deals are worth around $495 million and will help CrowdStrike's clients monitor, test, and cover AI models and agents.

Nimisha Asthagiri proposes a 'canvas' model for AI agents, providing bounded autonomy and continuous drift detection. This approach applies product engineering principles to AI development, enabling resilient software development.

A critical flaw in the EDK2 UEFI firmware, tracked as CVE-2026-97631, allows attackers to execute malicious code early in the boot process. The vulnerability is a memory-corruption flaw in an ICMP component and can be exploited to gain control over a server. It affects most server vendors and data centers, and patching is cumbersome and time-consuming.

ChatGPT claims to have achieved 20% of Google's search volume, with 1.2 billion weekly active users and 3.2 billion search-like queries. The company's models generate 1.5 billion images per week and 250 million people use ChatGPT Voice.

Google is offering schools the option to disable AI tools in the classroom, but this also means giving up Google Search. The company considers search and AI to be deeply integrated, and schools must choose between allowing all AI tools or none at all.

Fal, an AI inference startup, has raised $1.3 billion in its latest funding round, doubling its valuation to $16 billion. The company hosts and runs AI video, image, and audio models, and has surpassed $800 million in annualized revenue. This surge in demand for inference services has led to significant growth for Fal and other companies in the industry.

A company has implemented an automated code review system using a self-hosted 27B model, which reviews every PR in their main project. The system uses a combination of deterministic and inference-based approaches to provide feedback within 8 minutes. The model catches functional and non-functional issues, and engineers have shifted their focus to higher-level topics.

Apple is expanding Siri AI's capabilities to third-party apps, allowing developers to integrate the voice assistant into their apps. This update is part of the iOS 27.3 Developer Preview and includes a new 'app skills' feature that enables developers to combine multiple tasks into a single action.

Apple will enforce stricter age requirements on its App Store, focusing on dating apps and virtual companions, with apps featuring sexual or romantic content receiving a minimum age rating of 18+.

The AI industry has spent record sums on super PACs to elect friendly candidates, despite some executives calling for federal industry standards on AI. However, the public is souring on data centers, and the industry's massive donations may backfire if a newly empowered Congress imposes unwanted oversight. The industry needs to focus on problem-solving and addressing concerns about AI's impact on society.

Google will disable AI tools for K-12 students unless school administrators explicitly turn them on. However, schools that restrict AI altogether will no longer have access to Google Search, as AI features are now integrated into the search engine.

Google introduces LiteRT Tensor API, a lightweight C++ API for building, customizing, and optimizing machine learning models for on-device execution, providing flexibility and cross-platform portability.

Large language model hijacking attacks, known as LLMjacking, are causing significant financial losses for organizations, with some losing over $100,000 per day. Google's Threat Intelligence Group and Mandiant have identified two distinct operational clusters coordinating to compromise and monetize enterprise LLM infrastructure. Attackers gain access via exposed or compromised cloud credentials, then sell access to stolen API keys, resulting in fraudulent cloud charges.

Sonos is introducing a new voice assistant, Sonos 27voice, available in early access on November 6. The assistant can control playback, set timers, and answer general knowledge questions. It will be available to test in the US, in English, for users with Sonos 27 and S2 platform hardware.

North Korean hackers used social engineering tactics, including fake job offers, to trick a Binance employee into downloading malware, resulting in a record $6.3 billion heist. Researchers had previously warned about these tactics, which have been used to target the crypto industry and high-net worth individuals.

Sonos is rolling out a big app update that includes new Presets and early access to its AI voice assistant. The update allows users to create custom Presets to play music at a specific volume in a certain room. The Sonos 27voice AI beta will start on November 12th, enabling users to control their speakers with voice commands and access more features.

Binance reports a $6.3 billion crypto heist due to a fake job interview that led to malware installation, allowing North Korean hackers to compromise the company's wallet infrastructure.

Google has introduced a new top-tier reward of $250,000 for high-impact architectural security flaws in its Cloud Vulnerability Reward Program (VRP). The reward is for 'Root-in-Cloud' bugs, including tenant-isolation bypasses and complex cross-service logic flaws that allow attackers to gain full administrative control over arbitrary Google Cloud organizations.