Metomic

Two thirds of enterprises expect autonomous technology to fundamentally change how work gets done within two years. Fewer than one in four have the data strategy to support it. That gap is where breaches happen and where boards start asking uncomfortable questions. 93% of enterprises aren't ready. The number one reason? They don't know what sensitive data they have or where it lives. That's not a technology problem. That's a governance problem. A Harvard/Cloudera study found that protecting sensitive data and privacy is the top priority in enterprise data strategies, cited by 59% of respondents. Yet most organisations are still flying blind on what's sitting in their SaaS tools. You can't govern what you can't see. https://lnkd.in/eXRTMMmU

Slack retention settings tend to live in the admin panel and never get revisited. That's a problem, because every message you keep indefinitely is sensitive data your organisation is choosing to hold, and every channel with default settings is a choice you made by not making one. The practical issue is that most teams don't know what their plan actually allows. Slack's free tier retains only 90 days of messages; paid plans retain indefinitely by default. This comes up as a massive blocker when you're trying to control where customer data, HR conversations, or API keys end up sitting. The carousel below covers how retention works across Slack plans, where to set controls at workspace versus channel level, and how to remove sensitive content from specific messages without wiping entire conversation threads. Read the full piece on the Metomic Resource Centre: https://lnkd.in/gkRWxWXa Any other tips you'd share with the community? #SlackDataRetention

We’re heading to the Cloud & Cyber Security Expo on 4-5 March at Excel London, and we’re looking forward to the conversations already! As AI shifts from experimentation to everyday deployment, security teams need clarity on where sensitive data lives and how it’s being used. Metomic helps organisations stay in control across SaaS and cloud environments, with the visibility and guardrails required to roll out AI securely and confidently. If you’re attending and mapping out your AI plans for 2026, come and find us at the Nayaka stand. Send us a message to book time with the team, or stop by for a focused conversation about enabling AI without increasing risk. See you there! #CloudSecurity #CyberSecurity #AISecurity #DataSecurity #SaaSSecurity #TechShowLondon

Last year, security researchers demonstrated that Google's Gemini AI assistant could be manipulated through hidden prompt injection embedded inside a Google Calendar invite. Google has since mitigated the vulnerability. In the proof of concept, Gemini was asked to summarise a user's schedule and followed concealed instructions placed in the event description, writing the victim's private meeting data into a new calendar event the attacker could access. The technique required no malware and no account takeover. It relied entirely on how AI assistants interpret content inside trusted SaaS tools. This points us to a broader shift in enterprise risk. When AI systems are deeply integrated into platforms such as Google Calendar, Drive and Slack, they operate on top of existing data and permission structures. If sensitive information has been broadly shared over time, an injected instruction can simply guide the model towards material already within reach. Years of inherited permissions and forgotten links suddenly carry far greater weight. AI amplifies whatever sits underneath. At Metomic, we help teams gain visibility into sensitive data across SaaS environments, remediate overexposure at scale, and roll out AI safely. Click here to book a demo with us. https://lnkd.in/dRz9NtnK

Hi there! If you’re working through how to bring AI into your company, our CTO, Ben van Enckevort, is opening up time for informal 1:1 conversations. Whether you’re exploring agents, thinking about data access and governance, or just figuring out next steps, he’s happy to share what he’s learned from real-world deployments and talk through what you're working on. You can reach him via LinkedIn or at ben@metomic.io. Book your 1:1 now! To read more, click here: https://lnkd.in/ek2icKaU

OpenClaw has taken off quickly and it is easy to see why. The idea of a local AI agent that can actually take action on your behalf is compelling. What is less obvious is the level of access it typically runs with. API keys, bot tokens, OAuth secrets, filesystem permissions and shell access. Researchers have already found almost 5,000 exposed gateways based on the latest Shodan search, many without authentication. Investigations have also identified over 230 malicious skills reported live on ClawHub. That does not make OpenClaw risky by default, but it does mean configuration really matters. If you are evaluating it, or your team is already experimenting, it is worth understanding exactly what you are giving it access to and where the real risks sit. Read more here. https://lnkd.in/eWMZF_gN

AI tools are moving fast, but do we always know what access we’re giving them? Our clients at Metomic have shared a great blog on OpenClaw, a viral AI agent that can get plugged into inboxes, calendars, files and more. The piece highlights an important point: many users don’t realise how deep that access can go - and the security and data risks that come with it. A timely reminder that as AI becomes more autonomous, governance and visibility matter just as much as innovation. Worth a read 👇 https://lnkd.in/eWMZF_gN #AI #DataSecurity #AIGovernance #Risk #OpenClaw Bethany Ayers, Amy Beer, Xenia S., Helen Goldberg, Anna Simpson, Sally Blake

I spoke with Russ for Chief Executive Magazine about how we're building Metomic around AI employees. By the end of Q1, every one of our 25 people will manage at least four AI employees. We have a talented team and want to keep that density at a high speed. The old model of scaling by hiring has limits, every time you add a person to the team it adds friction. This approach lets us grow capacity without losing what makes the team great. We have to articulate what we want, why it matters, and what success looks like. The better we teach them, the more valuable they become. Calling them employees rather than tools is intentional, and it changes how people think about working with them. Which is why we had another 2-day session with Charlie Cowan a week ago. Sometimes it feels indulgent to block out time to learn and experiment with technology, to figure out ways to make our jobs less repetitive and more interesting. The article is worth a read if you're thinking about where AI fits in your own organisation. 

What's hiding behind the friendly lobster? 🦞 The autonomous agent that promises to take care of the mundane, and 'actually do things' has been everywhere lately. Under the friendly AI assistant managing your inbox, most people don't know that it routinely runs with access to API keys, bot tokens, OAuth secrets, filesystem permissions, and sometimes root-level execution inside containers. Want to try it for yourself? First, understand what you're giving it access to. Get the full guide with the 12 practical tips on how to use OpenClaw without getting burned > https://bit.ly/closeclaw #AIAgents #OpenClaw #ClawdBot #AIGovernance

If OpenClaw has been all over your feed the last two weeks, it's been all over your employees' feeds too. Some of them might have already installed it, connected it to email or their calendars. Maybe even their bank accounts. The promised productivity gains are tempting but there are the huge risks they don't know about. Security researchers have found 42,000+ exposed instances, 340+ malicious skills in the official marketplace already, and commodity infostealers are targeting OpenClaw config files. One researcher extracted a crypto private key via prompt injection in under five minutes. We wrote a technical briefing that covers: - What CISOs Need to Know This Week: The current threat landscape, sourced and timestamped - How OpenClaw Actually Works (In Risk Terms) - Architecture and where the gaps are - The "Obedient Monkey" Model, a framework that lands with non-technical people - Core Threats, such as prompt injection, data exfiltration, account takeover, supply chain risk - Where the Biggest Exposures Are Today: Shadow agents, exposed gateways, and over-permissive tools - Hardening Priorities: Policy, architecture, identity, monitoring - How to Talk About This With Your Board ("AI agent security" isn't a phrase that resonates in the boardroom) Read the full CISO Briefing → https://lnkd.in/gGPD4j6N