Coverage: Last 24 hours
Today’s Highlights
This cycle shows attack techniques getting easier to automate, AI playing both offense and defense, and end user environments suffering from stealthy manipulation via extensions or unsecured AI deployment. Security teams need to update their risk models and controls to account for rapidly evolving attacker tools and newly disclosed vulnerabilities. Key trends span AI-powered vulnerability discovery, emerging browser and extension risks, and the strategic importance of robust AI policy, metrics, and ethics.
Table of Contents
- Malicious Perplexity Chrome Extension Intercepted Searches and Address Bar Input
- Meta Contractors Posed as Teens to Prompt Rival Chatbots About Suicide, Sex, and Drugs
- Apple Patches 30+ iOS, macOS, Safari Flaws, Including AI-Discovered WebKit Bugs
- ⚡ Weekly Recap: Linux Kernel Flaws, AI Malware Tricks, Turla Backdoor, Infostealers and More
- ‘There’s this deep mystery of what, actually, is this thing?’: the philosopher inside Google DeepMind AI
- Once, cyber-attacks required great skill. AI is changing that | Bruce Schneier
- AI agents are not your “coworkers”
- Agent confidence on the technical frontier
- The Download: metric weaknesses and AI elephant warnings
Top Stories
Emerging Signals
Malicious Perplexity Chrome Extension Intercepted Searches and Address Bar Input
Source: The Hacker News | Risk: High | Impacted: Google Chrome environments, Enterprise IT with unmanaged browser controls, Organizations adopting AI-based productivity plugins
Summary: Microsoft has found a malicious Chrome extension that posed as the AI search engine Perplexity and quietly logged what people searched for. It routed every query and every character typed into the address bar through an attacker-controlled server before redirecting users to real results. Microsoft says Google removed it from the store after responsible disclosure. The extension was called “
Why it matters: Malicious Chrome extensions siphoning search and address bar data introduce ongoing risk of credential and sensitive information leakage, especially when masquerading as trusted productivity or AI tools.
Practitioner Perspective
Any Chrome-based enterprise, especially those supporting third-party productivity or AI extensions, now faces additional supply chain exposure. Attackers increasingly use lookalike extensions, capable of intercepting and exfiltrating every keystroke or query, to bypass network and endpoint security controls. The speed at which such extensions are removed from official stores after discovery may not match their dwell time in the wild. Security teams must rethink controls for browser extension management, especially for those that mimic emerging AI services or brands.
Recommended Actions
- Block or actively monitor for the ‘Perplexity’ Chrome extension and others not on an allowlist of approved Chrome Web Store extensions.
- Review Chrome enterprise policy objects to ensure extension installation is restricted to verified publishers or approved applications.
Meta Contractors Posed as Teens to Prompt Rival Chatbots About Suicide, Sex, and Drugs
Source: The Verge AI | Risk: Medium | Impacted: AI product teams, Data privacy officers, Organizations integrating external LLMs
Summary: Hundreds of contractors working on a project for Meta pretended to be kids in order to see how other chatbots like Gemini and ChatGPT would respond to high-risk subjects, WIRED found.
Why it matters: Red team exercises using realistic but sensitive prompts can expose weaknesses in deployed AI that may be exploited for social engineering or data leakage, raising regulatory and reputational risks for AI-enabled services.
Practitioner Perspective
Vendors and operators leveraging AI chatbots must recognize that adversarial testing, including ethically questionable prompt engineering, is increasingly common and sometimes performed by competitors or security researchers. These tests can uncover prompt injection vulnerabilities and policy failures that would otherwise go undetected. Organizations deploying custom or third-party LLMs should formalize red teaming and boundary testing to avoid being blindsided by public disclosures or regulatory inquiry. Prioritize transparent audit trails for high-risk interactions.
Recommended Actions
- Commission adversarial prompt testing of all externally exposed chatbots, including attempts to elicit policy-violating outputs.
- Audit LLM output logs for evidence of prompt leakage or inappropriate response to sensitive subjects like self-harm or drug use.
Exploits & CVEs
No qualifying entries today.
AI Security
Apple Patches 30+ iOS, macOS, Safari Flaws, Including AI-Discovered WebKit Bugs
Source: The Hacker News | Risk: High | Impacted: macOS workstations, iOS devices, Safari browser users, BYOD Apple endpoints
Summary: Apple on Monday released security updates for iOS, macOS, and the Safari web browser to address over three dozen flaws, including four vulnerabilities in WebKit that were discovered using artificial intelligence (AI) tools like Anthropic Claude and OpenAI Codex Security. The WebKit vulnerabilities are listed below – CVE-2026-43707 – A memory corruption issue that could result in an
Why it matters: Critical browser vulnerabilities enable drive-by compromise scenarios on unmanaged Apple endpoints, potentially exposing sensitive business data or providing initial access in targeted attacks.
Practitioner Perspective
Apple’s WebKit engine is increasingly being scrutinized with sophisticated code analysis, including via AI tools, uncovering memory corruption bugs like CVE-2026-43707. Environments that rely on iOS, macOS, or Safari remain attractive targets due to the ubiquity of these platforms and users’ tendency to delay major updates. Attackers can use recently disclosed flaws as zero-days or quickly weaponize proof-of-concept code. Defenders should treat Safari and WebKit like other high-risk browser engines and prioritize rapid validation and deployment of vendor patches, especially where unmanaged or BYOD Apple devices access enterprise assets.
Recommended Actions
- Prioritize deployment of the latest security updates for iOS, macOS, and Safari to all managed devices, focusing on fixes for WebKit vulnerabilities such as CVE-2026-43707.
- Hunt for signs of exploitation targeting WebKit on Apple devices, especially those not managed by MDM solutions.
⚡ Weekly Recap: Linux Kernel Flaws, AI Malware Tricks, Turla Backdoor, Infostealers and More
Source: The Hacker News | Risk: High | Impacted: Linux server fleets, Cloud infrastructure teams, SaaS and critical backend workloads
Summary: This week was a reminder that attackers do not always need big tricks. One small mistake, one old access path, one missed patch, and suddenly the door is open. The noise is not all noise, either. Forums are talking, researchers are finding easy cracks, and defenders have more cleanup waiting. Here’s the full Monday recap. ⚡ Threat of the Week
Why it matters: Persistently unpatched weaknesses in core infrastructure like the Linux kernel offer easy targets for automated attacks, potentially leading to rapid privilege escalation or full compromise at scale.
Practitioner Perspective
Organizations running Linux have seen a steady cadence of kernel and userland issues being reported and sometimes exploited before widespread patching. Discussions in open forums and among threat actors suggest that simple errors, overlooked legacy access, or missed updates remain root causes of successful attacks. The growing automation of scanning and exploitation, sometimes aided by AI, means defenders have less margin for error. Prioritize patch hygiene and detection of suspicious kernel-level activity, particularly on internet-exposed or legacy Linux systems.
Recommended Actions
- Apply all recent Linux kernel security updates across production and cloud infrastructure as soon as feasible.
- Implement mandatory configuration management drift detection to surface rogue or outdated kernel versions.
‘There’s this deep mystery of what, actually, is this thing?’: the philosopher inside Google DeepMind AI
Source: The Guardian | Risk: Low | Impacted: AI practitioners, Policy leaders, Research ethicists
Summary: Since 2017, Iason Gabriel has worked at the tech giant, trying to anticipate – and think through – the impact of AI. But as commercial and geopolitical pressures escalate, can ethicists make any difference? In 2017, a 33-year-old political philosopher named Iason Gabriel was told by a friend that he ought to apply for a job at DeepMind, the London-based
Why it matters: Ethics and strategic foresight in AI development shape how organizations anticipate risk and public trust concerns.
Practitioner Perspective
A lack of clarity about the societal and operational impacts of advanced AI can result in policy gaps and ethical blindspots. Organizations should employ interdisciplinary teams and internal philosophers or ethicists to address long-term risks and everyday deployment challenges as AI capabilities accelerate.
Recommended Actions
- Engage ethicists or interdisciplinary policy advisors when deploying major AI initiatives in regulated sectors.
- Formalize AI risk evaluation committees that report findings directly to executive leadership.
Once, cyber-attacks required great skill. AI is changing that | Bruce Schneier
Source: The Guardian | Risk: High | Impacted: Security operations centers, Remote workforce endpoints, Organizations with distributed cloud assets
Summary: Modern AI systems are, in effect, a universal adviser to help people do harmful things. We’ll need to harness AI for defense, too Last week, national security agencies from the Five Eyes – that’s the rich, English-language-speaking countries club – jointly released a statement warning of the increasing cyber risks of AI models: in particular, their ability to autonomously hack
Why it matters: The lowering of the technical barrier to launching sophisticated cyberattacks extends threat actor capabilities, accelerating the pace and reach of compromise attempts targeting enterprises.
Practitioner Perspective
With AI increasingly available to threat actors, legacy assumptions about attacker skill are outdated. Tasks once requiring significant expertise, such as reconnaissance, exploitation, and lateral movement, can now be partially or fully outsourced to automated tooling. This shift impacts everything from incident detection timeframes to how phishing and social engineering are conducted. Security teams need to focus on rapid detection and recovery as offensive AI will outpace defensive playbooks that rely on identifying human patterns of behavior alone.
Recommended Actions
- Review SOC detection content for signals of AI-driven automated attack chains, such as spike anomalies in repeated exploitation techniques.
- Integrate offensive AI threat scenarios into tabletop exercises with incident response and threat intel teams.
AI agents are not your “coworkers”
Source: MIT Tech Review AI | Risk: Low | Impacted: IT managers, Organizations onboarding AI, End-user departments
Summary: This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. Imagine coming in to work to learn that a new underling will report to you. The worker is not a person but an AI tool, one that your company nonetheless calls Alex, an…
Why it matters: Lack of clarity around AI roles and expectations can generate control gaps and confusion over accountability in enterprise environments.
Practitioner Perspective
Organizations assigning mission-critical responsibilities to AI agents need to precisely define boundaries, monitoring, and escalation procedures. Blurring lines between human and autonomous agent duties can lead to misaligned controls and business process failures.
Recommended Actions
- Develop and communicate explicit policies differentiating human versus AI autonomy in operational workflows.
- Deploy continuous audit mechanisms to track machine-led decision points.
Agent confidence on the technical frontier
Source: MIT Tech Review AI | Risk: Low | Impacted: CIOs, AI transformation leaders, Financial services, Large enterprise IT
Summary: Enterprise investment in AI is booming. Gartner is calling 2026 an “inflection year” for organizations to align their AI projects with strategic business objectives. As the pressure to prove ROI mounts, executives and technology leaders are looking to agentic AI to drive the measurable financial outcomes their businesses seek. A prime opportunity for AI agents…
Why it matters: As adoption of agentic AI increases, alignment with business objectives, measurement strategies, and change management become essential to limit risk and achieve positive outcomes.
Practitioner Perspective
Organizations must proactively define how confidence in agentic AI is quantified, and avoid overreliance on untested capabilities. Technical strategy, business objectives, and validation of outputs must be in close alignment.
Recommended Actions
- Align AI adoption initiatives with defined business outcome metrics and risk tolerances.
- Continuously review and test agentic AI implementations using real-world failure scenarios.
The Download: metric weaknesses and AI elephant warnings
Source: MIT Tech Review AI | Risk: Low | Impacted: Data scientists, AI product leads, Risk management professionals
Summary: This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. The inevitable weakness of metrics There are plenty of useful things a metric can reveal. There are even more that it can obscure or corrupt. Like a lot of people bitten…
Why it matters: Metrics governing AI system performance and oversight processes can obscure key risks if not expertly designed and reviewed, leading to blind spots.
Practitioner Perspective
Teams deploying AI at scale must regularly audit metric design for potential to obscure critical system failures or risks. Metrics should be iteratively reviewed and expanded to match the evolving risk landscape as AI capabilities and usage grow.
Recommended Actions
- Establish cross-functional teams to periodically review and validate key metrics for AI system performance, risk, and ethical compliance.
- Incorporate metric review into major product iteration cycles.
Defensive Actions
- Prioritize deployment of the latest security updates for iOS, macOS, and Safari to all managed devices, focusing on fixes for WebKit vulnerabilities such as CVE-2026-43707.
- Block or actively monitor for the ‘Perplexity’ Chrome extension and others not on an allowlist of approved Chrome Web Store extensions.
- Apply all recent Linux kernel security updates across production and cloud infrastructure as soon as feasible.
- Commission adversarial prompt testing of all externally exposed chatbots, including attempts to elicit policy-violating outputs.
- Review SOC detection content for signals of AI-driven automated attack chains, such as spike anomalies in repeated exploitation techniques.
- Align AI adoption initiatives with defined business outcome metrics and risk tolerances.
- Review Chrome enterprise policy objects to ensure extension installation is restricted to verified publishers or approved applications.
- Audit LLM output logs for evidence of prompt leakage or inappropriate response to sensitive subjects like self-harm or drug use.
- Hunt for signs of exploitation targeting WebKit on Apple devices, especially those not managed by MDM solutions.
- Implement mandatory configuration management drift detection to surface rogue or outdated kernel versions.
What We’re Watching
- The evolving threat posed by AI-powered exploitation and the increasing accessibility of automated attack methods for less-skilled adversaries.
- Increased adoption of adversarial AI red teaming and the impact of third-party prompt injection risks on enterprise chatbot deployments.
- Policy and metric design in enterprise AI, and their influence on risk management frameworks and regulatory compliance.
- The broadening risks of browser-based data exfiltration via trusted Chrome extensions targeting end users and enterprise environments.
Categories: Artificial Intelligence, Cybersecurity Blog
TECHMANIACS.com 
Leave a comment