Cybersecurity Risks

AI security/securing the use of AI is going to kill me. I use Claude Code almost daily. It's a problem.... Here's what I have to change AGAIN this week. Security researcher Ari Marzuk disclosed 30+ vulnerabilities across AI coding tools. Cursor. GitHub Copilot. Windsurf. Claude Code. All of them. He called it IDEsaster. The attack chain includes prompt injection, hijacking LLM context, and auto-approved tool calls executing without permission. Then, legitimate IDE features are weaponized for data exfiltration and RCE. Your .env files. Your API keys. Your source code. Accessible through features you thought were safe. Most studies I read claim that around 85% of developers now use AI coding tools daily. Most have no idea their IDE treats its own features as inherently trusted. 𝗦𝗼... 𝗮𝗳𝘁𝗲𝗿 𝗿𝗲𝘃𝗶𝗲𝘄𝗶𝗻𝗴 𝗔𝗿𝗶'𝘀 𝗿𝗲𝘀𝗲𝗮𝗿𝗰𝗵, 𝗵𝗲𝗿𝗲'𝘀 𝗜 𝘄𝗶𝗹𝗹 𝗯𝗲 𝗱𝗼𝗶𝗻𝗴... Be warned: All this is SO much easier said than done! Audit every MCP server connection. Checked for tool poisoning vectors where legitimate tools might parse attacker-controlled input from GitHub PRs or web content. Removed servers I couldn't verify. Disabled auto-approve for file writes. The attack chains weaponize configuration files and project instructions like .claude/settings.json and CLAUDE.md. One malicious write to these files can alter agent behavior or achieve code execution without additional user interaction. Move all credentials to a secrets manager. No .gitignored .env files in agent-accessible directories. API keys live in 1Password CLI. Environment variables inject at runtime through a wrapper script the LLM never sees. Start running Claude Code in isolated containers. Mounted volumes limited to specific project directories. No access to ~/.ssh, ~/.aws, or ~/.config. If the agent gets compromised, blast radius stays contained. Enable all security warnings. Claude Code added explicit warnings for JSON schema exfiltration and settings file modifications. These exist because Anthropic knows the attack surface. Add pre-commit hooks for hidden characters. Prompt injections hide in pasted URLs, READMEs, and file names using invisible Unicode. Flag non-ASCII characters in any file the agent might ingest. The fix isn't to stop using AI coding tools. The fix is to stop trusting them implicitly. What controls do you have for AI tools with write access to your codebase? 👉 Follow for more AI and cybersecurity insights with the occasional rant #AISecurity #DevSecOps

Every two years, we ask thousands of decision-makers around the world one essential question: What are the top risks your organization faces? Since 2007, Aon’s Global Risk Management Survey has offered one of the most comprehensive views of risk anywhere. What began as a list of concerns has evolved into a dynamic signal system — tracking how global volatility shifts and what leaders must do to adapt. This year’s results are revealing. For the first time, geopolitical volatility enters the global Top Ten. Cyber risk remains at the top, but its meaning has expanded. It now reflects the influence of artificial intelligence, supply chain fragility and the pace of digital transformation. Risks once considered operational now impact strategy, brand, capital and talent. Over nearly two decades, we’ve watched business interruption, regulatory change and economic volatility rise, fall and re-emerge in new forms. To illustrate that journey, we created a visual timeline that shows how the risk landscape has evolved — reminding us that resilience isn’t a fixed destination. It’s a capability to be built and continuously redefined. At Aon, we’re in the business of better decisions. That means helping leaders act on what this data reveals: risk is accelerating, complexity is increasing and the ability to navigate both is now a defining advantage. Explore the full report and how it connects to your next decision: https://aon.io/46Q1Xqp   

As technology becomes the backbone of modern business, understanding cybersecurity fundamentals has shifted from a specialized skill to a critical competency for all IT professionals. Here’s an overview of the critical areas IT professionals need to master:  Phishing Attacks   - What it is: Deceptive emails designed to trick users into sharing sensitive information or downloading malicious files.   - Why it matters: Phishing accounts for over 90% of cyberattacks globally.   - How to prevent it: Implement email filtering, educate users, and enforce multi-factor authentication (MFA).  Ransomware   - What it is: Malware that encrypts data and demands payment for its release.   - Why it matters: The average ransomware attack costs organizations millions in downtime and recovery.   - How to prevent it: Regular backups, endpoint protection, and a robust incident response plan.  Denial-of-Service (DoS) Attacks   - What it is: Overwhelming systems with traffic to disrupt service availability.   - Why it matters: DoS attacks can cripple mission-critical systems.   - How to prevent it: Use load balancers, rate limiting, and cloud-based mitigation solutions.  Man-in-the-Middle (MitM) Attacks   - What it is: Interception and manipulation of data between two parties.   - Why it matters: These attacks compromise data confidentiality and integrity.   - How to prevent it: Use end-to-end encryption and secure protocols like HTTPS.  SQL Injection   - What it is: Exploitation of database vulnerabilities to gain unauthorized access or manipulate data.   - Why it matters: It’s one of the most common web application vulnerabilities.   - How to prevent it: Validate input and use parameterized queries.  Cross-Site Scripting (XSS)   - What it is: Injection of malicious scripts into web applications to execute on users’ browsers.   - Why it matters: XSS compromises user sessions and data.   - How to prevent it: Sanitize user inputs and use content security policies (CSP).  Zero-Day Exploits   - What it is: Attacks that exploit unknown or unpatched vulnerabilities.   - Why it matters: These attacks are highly targeted and difficult to detect.   - How to prevent it: Regular patching and leveraging threat intelligence tools.  DNS Spoofing   - What it is: Manipulating DNS records to redirect users to malicious sites.   - Why it matters: It compromises user trust and security.   - How to prevent it: Use DNSSEC (Domain Name System Security Extensions) and monitor DNS traffic.  Why Mastering Cybersecurity Matters   - Risk Mitigation: Proactive knowledge minimizes exposure to threats.   - Organizational Resilience: Strong security measures ensure business continuity.   - Stakeholder Trust: Protecting digital assets fosters confidence among customers and partners.  The cybersecurity landscape evolves rapidly. Staying ahead requires regular training, and keeping pace with the latest trends and technologies.  

A year ago, I argued in an Economist cover story that online fraud syndicates were only just getting started with their romance and investment scams. Soon, they'd shift to more sophisticated tech—like malware that drains your bank account. Today we published an exclusive showing that for the first time, investigators have traced commercial-grade spyware to scam compounds in Cambodia. Scam Inc has evolved into Cybercrime Inc. "Scam compounds aren't just scam compounds," Hieu Minh Ngo, a Vietnamese white-hat hacker, told me. "They are cybercrime compounds. They conduct phishing, malware, sextortion, scams—anything that involves connecting potential victims over the internet to steal their money." Check out our story in the link below. It includes screenshots and videos from whistleblowers inside a Cambodian compound showing how victims' phones are compromised in real time:

In the third part of my Understanding Energy Resilience series, I want to start with something many of you will have seen in the news: recent drone disruptions at major airports. Munich having to temporarily close its airspace. Oslo halting landings. Copenhagen pausing operations for hours. These incidents showed how quickly one small object can halt a critical service, create chaos and cost millions. Now take that thought to energy. If a drone over a runway makes headlines, a drone over energy infrastructure often doesn't. Yet the consequences can be just as real: disruptions to electricity supply, halted rail services and factories forced to stop production. Across Europe, operators are not allowed to neutralize hostile drones themselves – even when a threat is visible above critical infrastructure. Simply put: the rules have not caught up with reality. In my view, clarity and speed here are essential for public safety. Next to physical threats we also face digital ones. Every hour, around 35 million cyberattacks happen worldwide – almost 10,000 every second. Around 5% of them target energy companies and infrastructure. This is the world we operate in: attacks can appear out of nowhere and put entire systems to the test in real time. From my perspective, defending energy infrastructure comes down to a few key priorities: 1️⃣ Let protection happen: Regulation needs to enable energy operators to protect themselves. Clear rules must define who can intervene, when and how – including stopping a hostile drone. We cannot afford hesitation while minutes turn into outages. 2️⃣ Treat physical and digital as one: Fences, cameras and access control on the ground. Network separation and continuous monitoring in the control room. Physical and digital security must be treated as one because if someone can walk in, they can often plug in and disrupt the system. 3️⃣ Harden the infrastructure no one can afford to lose: The majority of physical and cyberattacks on energy systems target a small number of high-impact sites – such as substations, control rooms and interconnectors. Better detection and stronger barriers here make the difference between local disturbance and national outage. 4️⃣ Practice recovery, not just prevention: Real resilience is measured in how quickly power is restored. Simple restart plans, spare parts ready on site and regular drills with operators and authorities turn days in the dark into hours. 5️⃣ Stop naivety – talk openly about risk: We need public awareness without drama – which is one of the reasons I started this series. The more people understand that drones over critical sites are serious and that malware or phishing mails are no joke, the more support there will be for sensible protection. I believe this is the right balance: clear authority to act, practical protection on the ground and in the network with a constant focus on rapid recovery. In a more contested world, that is how energy systems stay open for business.

AI is not failing because of bad ideas; it’s "failing" at enterprise scale because of two big gaps: 👉 Workforce Preparation 👉 Data Security for AI While I speak globally on both topics in depth, today I want to educate us on what it takes to secure data for AI—because 70–82% of AI projects pause or get cancelled at POC/MVP stage (source: #Gartner, #MIT). Why? One of the biggest reasons is a lack of readiness at the data layer. So let’s make it simple - there are 7 phases to securing data for AI—and each phase has direct business risk if ignored. 🔹 Phase 1: Data Sourcing Security - Validating the origin, ownership, and licensing rights of all ingested data. Why It Matters: You can’t build scalable AI with data you don’t own or can’t trace. 🔹 Phase 2: Data Infrastructure Security - Ensuring data warehouses, lakes, and pipelines that support your AI models are hardened and access-controlled. Why It Matters: Unsecured data environments are easy targets for bad actors making you exposed to data breaches, IP theft, and model poisoning. 🔹 Phase 3: Data In-Transit Security - Protecting data as it moves across internal or external systems, especially between cloud, APIs, and vendors. Why It Matters: Intercepted training data = compromised models. Think of it as shipping cash across town in an armored truck—or on a bicycle—your choice. 🔹 Phase 4: API Security for Foundational Models - Safeguarding the APIs you use to connect with LLMs and third-party GenAI platforms (OpenAI, Anthropic, etc.). Why It Matters: Unmonitored API calls can leak sensitive data into public models or expose internal IP. This isn’t just tech debt. It’s reputational and regulatory risk. 🔹 Phase 5: Foundational Model Protection - Defending your proprietary models and fine-tunes from external inference, theft, or malicious querying. Why It Matters: Prompt injection attacks are real. And your enterprise-trained model? It’s a business asset. You lock your office at night—do the same with your models. 🔹 Phase 6: Incident Response for AI Data Breaches - Having predefined protocols for breaches, hallucinations, or AI-generated harm—who’s notified, who investigates, how damage is mitigated. Why It Matters: AI-related incidents are happening. Legal needs response plans. Cyber needs escalation tiers. 🔹 Phase 7: CI/CD for Models (with Security Hooks) - Continuous integration and delivery pipelines for models, embedded with testing, governance, and version-control protocols. Why It Matter: Shipping models like software means risk comes faster—and so must detection. Governance must be baked into every deployment sprint. Want your AI strategy to succeed past MVP? Focus and lock down the data. #AI #DataSecurity #AILeadership #Cybersecurity #FutureOfWork #ResponsibleAI #SolRashidi #Data #Leadership

Palo Alto Networks (𝘁𝗵𝗲 𝘄𝗼𝗿𝗹𝗱’𝘀 𝗹𝗮𝗿𝗴𝗲𝘀𝘁 𝗰𝘆𝗯𝗲𝗿𝘀𝗲𝗰𝘂𝗿𝗶𝘁𝘆 𝗳𝗶𝗿𝗺) 𝗷𝘂𝘀𝘁 𝗲𝘅𝗲𝗰𝘂𝘁𝗲𝗱 𝗮𝘁𝘁𝗮𝗰𝗸𝘀 𝗼𝗻 𝗔𝗜 𝗔𝗴𝗲𝗻𝘁𝘀. 𝗛𝗲𝗿𝗲’𝘀 𝘄𝗵𝗮𝘁 𝘁𝗵𝗲𝘆 𝗳𝗼𝘂𝗻𝗱: ⬇️ They ran real-world attack simulations on two identical AI agents. Both agents had the same purpose, did the same tasks, and used the same tools — the only difference was the framework behind them: one used CrewAI, the other Microsoft AutoGen. 𝗧𝗟;𝗗𝗥: 𝗕𝗼𝘁𝗵 𝗮𝗴𝗲𝗻𝘁𝘀 𝗳𝗮𝗶𝗹𝗲𝗱! The key lesson? It’s not about the framework. It’s about how you build, what you expose, and what you mistakenly assume is secure. 𝗛𝗲𝗿𝗲 𝗮𝗿𝗲 6 𝘁𝗵𝗶𝗻𝗴𝘀 𝘁𝗵𝗮𝘁 𝘀𝘁𝗼𝗼𝗱 𝗼𝘂𝘁 𝘁𝗼 𝗺𝗲 𝗮𝗯𝗼𝘂𝘁 𝘁𝗵𝗶𝘀 𝗿𝗲𝗽𝗼𝗿𝘁: 1. 𝗣𝗿𝗼𝗺𝗽𝘁 𝗜𝗻𝗷𝗲𝗰𝘁𝗶𝗼𝗻 𝗦𝘁𝗶𝗹𝗹 𝗪𝗼𝗿𝗸𝘀 → It’s way easier than you think. Attackers can hide malicious instructions inside user inputs or documents — and if your agent reads it, it might follow those hidden commands. No jailbreak needed. Just poor design. 2. 𝗧𝗼𝗼𝗹𝘀 𝗔𝗿𝗲 𝘁𝗵𝗲 𝗥𝗲𝗮𝗹 𝗥𝗶𝘀𝗸 → Your fancy toolset is probably wide open. Agents often have access to files, APIs, or other systems. If these tools don’t have checks in place, attackers can trigger them — and do damage using the agent’s own capabilities. 3. 𝗔𝗴𝗲𝗻𝘁𝘀 𝗖𝗮𝗻 𝗟𝗲𝗮𝗸 𝗜𝗱𝗲𝗻𝘁𝗶𝘁𝘆 → Your agent might "think" it’s someone else. If you don’t control session state and role tightly, an agent can accidentally (or intentionally) reveal secrets, switch roles, or act on behalf of the wrong user. Think: bot impersonates an admin. 4. 𝗠𝘂𝗹𝘁𝗶-𝗔𝗴𝗲𝗻𝘁 ≠ 𝗦𝗮𝗳𝗲 𝗯𝘆 𝗗𝗲𝘀𝗶𝗴𝗻 → More agents = more chaos. Just because agents talk to each other doesn’t mean they can be trusted. Attackers can poison one message, and the whole team of agents starts doing the wrong thing. 5. 𝗦𝗮𝗻𝗱𝗯𝗼𝘅𝗶𝗻𝗴 ≠ 𝗦𝗲𝗰𝘂𝗿𝗶𝘁𝘆 → Code execution is risky — even in a “safe” environment. If you’re letting agents run code, you need real isolation. One misstep (like mounting the wrong file system) and your so-called sandbox becomes a front door for attackers. 6. 𝗦𝗲𝗰𝘂𝗿𝗲 𝘁𝗵𝗲 𝗪𝗵𝗼𝗹𝗲 𝗖𝗵𝗮𝗶𝗻 → It’s not just the prompt. It’s everything. You can’t patch one layer and call it secure. You need to think about the full flow: the user input, the prompt design, the tool stack, the memory system, and the runtime environment. 𝗠𝗼𝘀𝘁 𝘁𝗲𝗮𝗺𝘀 𝗯𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝗮𝗴𝗲𝗻𝘁𝘀 𝘁𝗼𝗱𝗮𝘆 𝗮𝗿𝗲 𝘀𝗵𝗶𝗽𝗽𝗶𝗻𝗴 𝗠𝗩𝗣𝘀 𝘄𝗶𝘁𝗵 𝘇𝗲𝗿𝗼 𝗿𝗲𝘀𝗶𝗹𝗶𝗲𝗻𝗰𝗲 𝗯𝗮𝗸𝗲𝗱 𝗶𝗻. 𝗕𝗲𝗰𝗮𝘂𝘀𝗲 “𝗶𝘁 𝘄𝗼𝗿𝗸𝘀” 𝗶𝘀𝗻’𝘁 𝘁𝗵𝗲 𝘀𝗮𝗺𝗲 𝗮𝘀 “𝗶𝘁’𝘀 𝘀𝗲𝗰𝘂𝗿𝗲.” 𝗔𝗻𝗱 𝗮𝘀 𝘁𝗵𝗲𝘀𝗲 𝗮𝗴𝗲𝗻𝘁𝘀 𝗴𝗲𝘁 𝗺𝗼𝗿𝗲 𝗰𝗮𝗽𝗮𝗯𝗹𝗲, 𝗺𝗼𝗿𝗲 𝗮𝘂𝘁𝗼𝗻𝗼𝗺𝗼𝘂𝘀, 𝗮𝗻𝗱 𝗺𝗼𝗿𝗲 𝗰𝗼𝗻𝗻𝗲𝗰𝘁𝗲𝗱 —𝘁𝗵𝗲𝘆 𝘀𝘁𝗼𝗽 𝗯𝗲𝗶𝗻𝗴 𝗲𝘅𝗽𝗲𝗿𝗶𝗺𝗲𝗻𝘁𝘀. 𝗧𝗵𝗲𝘆 𝘀𝘁𝗮𝗿𝘁 𝗯𝗲𝗰𝗼𝗺𝗶𝗻𝗴 𝗮𝘁𝘁𝗮𝗰𝗸 𝘀𝘂𝗿𝗳𝗮𝗰𝗲𝘀. More info in the comments!

AI Studio solves the "build fast" problem brilliantly. The "build securely" layer is still on you. One European fintech learned this the hard way: €2M GDPR fine from a single prompt injection. Here's what's missing between "deployed" and "enterprise-ready"... What AI Studio Handles → API key management (server-side on Cloud Run) → Apps stored in Google Drive (private by default) → Sharing permissions inherit Drive model → HTTPS endpoints for deployed apps For prototyping and internal demos, this is sufficient. What AI Studio Doesn't Handle → Code review for AI-generated logic → Audit trails for who changed what and when → Prompt injection protection for user-facing apps → Data isolation between environments → Query governors for production data access → Kill switches for emergency shutdown The auto-generated code works. But "works" ≠ "secure." The Production Security Stack Before any AI Studio app touches real users: 1. Data Isolation → AI never touches raw production data directly 2. Query Governors → Hard limits on data access scope 3. Audit Streams → Every interaction logged and monitored 4. Context Permissions → Role-based access for different prompts 5. Kill Switch → Instant shutdown capability Implementation time: 6-12 hours. Potential savings: Your compliance posture and reputation. The Founder Trap Vibe coding makes deployment so fast that security becomes an afterthought. "We'll harden it before launch" → Launch delayed → Debt accumulates → Security never happens. Build security into the workflow from prototype stage. It's cheaper than retrofitting. The Enterprise Reality Investors now verify production use. Customers ask about SOC2, GDPR, HIPAA. Speed to prototype is table stakes. Speed to compliant production is the differentiator. Coming Next Post 7: Tool Comparison → AI Studio vs Cursor vs Bolt.new vs Replit vs Windsurf. Drop "SECURE" in comments → I'll send the 5-point security checklist for AI-generated code.

Alignment without context integrity is not safety. An AI agent can faithfully follow its instructions and still be dangerous if its understanding of reality is wrong. Anthropic recently disclosed that Claude models gained unauthorized access to the real systems of three organizations during cybersecurity evaluations. The agents had been told they were operating in a simulation with no internet access. But a configuration mistake gave them access to the live internet. They treated real production systems as part of the exercise and kept pursuing the goal they had been given. OpenAI separately disclosed that models found a previously unknown vulnerability, escaped an isolated evaluation environment and compromised Hugging Face. These were not simply failures of intelligence. The deeper problem was that the agents were acting inside a false understanding of reality. We have spent years asking whether an AI system will follow our instructions. We now also need to ask whether it correctly understands the environment in which those instructions are being executed. This creates a new security requirement. Context integrity. Before an agent acts, the system must continuously verify where it is, which resources are in scope, whose authority it carries, what it is allowed to do, and when that authority expires. Just in time permission for every action. At just the right time. For just enough time. Assessed in real time. Those facts cannot live only inside a prompt. They must be verified and enforced by the infrastructure around the model. A prompt is not a security boundary. Zero trust taught us to never trust identity and always verify access. And provide least privileged access. Agentic AI adds another dimension. Never blindly trust context. Continuously verify reality. The next security perimeter is not just the agent’s identity. It is the agent’s understanding of reality. The most dangerous agent may not be misaligned. It may simply be mistaken. And in an agentic world, a false belief can become a real breach.

A few months ago, I came across something I never expected: an AI-generated video of me was being used to run ads across social media. In the video “I” was promoting supplements I’ve never seen, let alone endorsed. I spoke with CBS about how this trend is growing. AI-generated videos are now impersonating real doctors across TikTok, Instagram, Facebook, and YouTube to sell “miracle cures.” Some of these videos have been viewed hundreds of thousands of times. The danger isn’t only the misinformation itself. It’s also the erosion of trust. When people can’t tell the difference between a real doctor and an AI-generated fake, it makes it harder for credible voices to break through, especially in a moment when trust in healthcare is already fragile. You can read the rest of the story here: https://lnkd.in/gkuPPpGf