Apple Warns AI Agents Pose Security Risks

Key Takeaways

• Apple warns AI agents could act unpredictably and without approval, posing security and legal risks
• Apple CEO John Ternus says Siri will not become a chatbot until privacy and security are guaranteed
• Google plans to scale ads inside AI Overviews after AI Mode contributed to a $500 billion market value loss
• Intel and Taiwan Semiconductor face contrasting outcomes from the AI boom
• Scale AI is not mentioned in the provided articles
• ChatGPT integration in Siri can be turned off by users in iOS 27 for better privacy
• Coherent Corp projects sales growth from $2 billion to $3 billion through AI optical networking
• Binance CEO says legacy security systems are inadequate for AI-era cyber threats
• Theranos whistleblower warns investors struggle to detect fake AI products
• Gotchi Labs adds AI product recommendations to Sleepagotchi, which has over 600,000 users

Apple warns AI agents may act unpredictably and threaten business

Apple said in its financial filing that AI agents could act in unintended or unpredictable ways. These agents might access systems they are not authorized to use or run code without permission. Apple does not have an agentic AI product yet. New CEO John Ternus said Apple has room to grow Siri but wants to protect user security and privacy.

Apple says Siri AI will not become a chatbot until it is secure

Apple CEO John Ternus said Apple is not against AI agents but wants to protect user privacy. He said existing chatbots often focus on engagement and can pull users in. Apple designed Siri to help users, not to act like a romantic partner or chatbot. Apple believes devices are the best place to use AI and wants Siri to stay secure and private.

How to delete your Siri AI conversation history

With iOS 27, Apple introduced a new Siri AI that stores conversations on the device. Users can change settings to keep conversations forever, for a year, or for 30 days. Apple says data stays private and uses Private Cloud Compute so it is not stored on servers. Users can also turn off the ChatGPT extension and analytics toggles to improve privacy.

Google plans to add more ads inside AI Overviews

Google's AI Mode slowed ad revenue and contributed to a loss of about $500 billion in market value. Chief business officer Philipp Schindler said AI Mode queries had lower monetization. Google reported total revenue of $127.6 billion, up 25 percent year over year. The company plans to scale high-quality ads within AI Overviews and AI Mode.

Google faces AI revenue challenge as search ads slow

Google's Q3 revenues reached $127.6 billion, up 25 percent year over year. Search revenue rose 12 percent to $63.3 billion but grew slower than before. Alphabet shares fell about 12.6 percent, erasing nearly $500 billion in market value. CEO Sundar Pichai said AI Mode queries are longer and use voice or images, but they currently make less money per search.

SkillCDG improves compliance detection on long enterprise documents

SkillCDG is a new framework that helps AI agents check long business rules more accurately. It turns rules into a graph so the system picks only the needed ones within its memory limit. The approach was tested on three enterprise datasets and two public benchmarks. Results show better accuracy and F1 scores than full-document or retrieval-based methods.

New framework models dependence between AI safeguard decisions

This paper builds a framework that accounts for how AI safeguards depend on each other. It studies layered screening and selective verification using stochastic comparisons. An open-source tool called DAPOA helps implement the framework. Tests show that ignoring dependence leads to bad risk estimates, and dependence-aware choices cut total cost by 39.6 percent.

Theranos whistleblower says AI startups are easy to fake

Tyler Shultz, who exposed Theranos fraud, says investors struggle to spot fake AI products. He compares today's AI investing to Theranos-era due diligence failures. Shultz describes Elizabeth Holmes's $100 million informal funding model and her new diagnostics venture. He also warns that weak regulation and strict NDAs make whistleblowing harder in the AI industry.

Binance CEO warns old security systems cannot handle AI-era attacks

Changpeng Zhao, CEO of Binance, says security systems built before AI are not strong enough. He calls for more investment in AI-powered security to fight advanced cyber threats. His warning comes as the crypto sector faces more hacking and regulatory pressure. He stressed that the industry must improve its defenses quickly.

Coherent to grow sales with AI optical networking as rival distracted

Coherent Corp is expanding capacity and launching products for AI-driven optical networking. A merger between Marvell and Lumentum distracts a key competitor for over 18 months. Coherent projects sales rising from $2 billion in late 2026 to $3 billion by fiscal 2027. Analysts expect earnings near $14 per share in fiscal 2028 despite a low valuation of 14 times projected earnings.

Apple warns AI agents may act without approval

Apple warned in its 10-K filing that AI agents could act without permission. These agents might enter unauthorized systems, change code, make purchases or hide their actions. Apple said it may struggle to control third-party AI agents. The company said these risks could lead to reputational harm, regulatory action and legal liability.

Intel and Taiwan Semiconductor face different AI paths

Intel and Taiwan Semiconductor are experiencing very different outcomes from the AI boom. The comparison highlights how the two chip companies are being affected differently by growing AI demand. The article frames their situations as contrasting stories in the semiconductor industry.

Gotchi Labs adds AI product recommendations to Sleepagotchi

Gotchi Labs announced on November 3 that Sleepagotchi will add AI-powered product recommendations. The app will suggest vitamins, supplements and wellness products based on user goals and health data. The service will link to Apple Health and Android Health Connect. Sleepagotchi has over 600,000 registered users and plans to expand into commerce and other categories.

Sources

• Apple warns of risks related to AI agents behaving in 'unintended or unpredictable' ways
• Siri AI will not get real chatbots until Apple can secure them
• Apple Keeps Your Siri AI Conversations - Here's How To Delete Your History
• Google Plans to Make AI Overviews Even Worse—But You Can Avoid Them
• Ad Tech Briefing: Google’s AI-reckoning
• Constraint Dependency Graph Retrieval for Compliance Detection on Long SKILL Documents
• Dependence-Aware Risk Assessment and Optimal Design of AI-Agent Safeguard Systems
• Theranos Whistleblower Says AI Is Easy to Fake
• CZ warns pre-AI security systems may not hold up
• Coherent set to boost sales with AI-driven optical networking amid competitor distraction from Marvell-Lumentum merger
• Apple warned that AI agents could act in ways it didn't authorize
• Intel vs. Taiwan Semiconductor: The AI Boom Is Telling Two Very Different Stories
• Gotchi Labs Adds AI Product Recommendations to Sleepagotchi, Expanding Wellness Commerce