Upgrade to Pro — share decks privately, control downloads, hide ads and more …

HYCU - IT Press Tour #69 September 2026

HYCU - IT Press Tour #69 September 2026

Avatar for The IT Press Tour

The IT Press Tour PRO

September 01, 2026

More Decks by The IT Press Tour

Other Decks in Technology

Transcript

  1. Globally Recognized and Trusted Key Investors Trusted by thousands of

    organizations in over 78 countries Strategic Alliances
  2. HYCU named a Leader in IDC MarketScape for Worldwide SaaS

    Data Protection 2025-2026 Vendor Assessment Leader in the GigaOM Cloud Native Data Protection Radar Report Visionary in Gartner Magic Quadrant for Enterprise Backup and Recovery
  3. Agenda 01 Welcome 02 Era of Agentic Risk 03 ITPT

    Exclusive: AI Visibility 04 Protection & Continuity 05 Lightning struck the backups! 06 Q&A
  4. WE TAL KED ABOUT THIS L AST YEAR… A new

    failure mode What we are hearing from customers, and why Europe is asking first. Tadej Pogačar
  5. THE NEW FAILURE MODE A third actor is inside the

    perimeter Agent Already inside 00:00 9 seconds Time an AI agent took to delete a company's production database and all its backups. PocketOS, April 2026. Human error inside Ransomware outside
  6. THE NEW FAILURE MODE Three ways an agent destroys data

    Drift Mis instruction Standing credentials The agent hallucinates the state of a system and acts on a reality that does not exist. A hidden or injected instruction redirects the agent toward destruction. A hidden or injected instruction redirects the agent toward destruction. Gemini CLI, July 2025. Agent invented a directory, moved files into it, overwrote the originals. Amazon Q for VS Code, July 2025. Injected prompt ordered the agent to wipe the filesystem and cloud resources. Shipped to more than 964,000 installs. PocketOS, April 2026. One API token with full access dropped the production database and its backups.
  7. THE NEW FAILURE MODE Your systems of record were built

    for human mistakes 93 90 60 15 9 SharePoint Okta audit log Jira Salesforce seconds agent action When prevention fails at machine speed, the ability to recover is the only control left.
  8. How Atlassian’s Rovo can permanently modify or delete your records

    The example: an admin asks Rovo, “Archive all resolved tickets older than 90 days in the SUP project.” A reasonable request, with an irreversible outcome. 1 · Instruction 2 · Acts as you 3 · Machine speed 4 · Atlassian commits 5 · Point of no return Rovo reads the request and plans a chain of actions to satisfy it. It runs with a service account's permissions. No second human approves. It reads “archive” as bulk delete plus status change, and fires hundreds of REST API calls in seconds. Each call is a valid, authorized write. Issues are deleted; SLA timers and fields are overwritten. Overwrites have no undo. Deletions land in a short-lived trash. Noticed days later.
  9. FIRS T L OOK FROM IT P RES S TOU

    R Agent visibility and usage have never been more challenging Postojna Cave Park
  10. Protect at the speed of agents. Stay in business during

    the incident. Ask your backups the question. See the whole board. Hold every operator to account.
  11. A repository is no longer just a code repo Application

    code Runbooks Infrastructure as code SDKs CI/CD pipelines Deployment configs Repository System of record AI prompts Agent definitions Losing a repo can mean losing the specification for how the company operates.
  12. Even before AI, there were many ways to permanently lose

    GitHub data 30+ Accidental deletion & force-push Insider & offboarding git rm, git push --force, rebase and history rewrites Departing employees, lingering access, deliberate sabotage Credential & token compromise Misconfigured automation Phishing, leaked PATs and SSH keys, account takeover CI/CD scripts, over-privileged Actions, scheduled jobs HYCU catalogued more than 30 ways repo data gets deleted, corrupted, or altered. None are recoverable by the platform’s own resilience. “To err is human. To delete main, developer.” Provider & integration failure Outages, third-party OAuth abuse, mirroring mistakes
  13. New Failure Modes in GitHub A cleanup agent deletes branches

    that looked stale but were not. A refactor agent force-pushes to main after a bad rebase. A migration script rewrites history across a dozen repos at once. A compromised agent token deletes branches, tags, or releases before anyone notices. A CI workflow runs a destructive git command an agent generated. In every case the damage looks legitimate to GitHub. It was authenticated. It was authorized. Native protection has no way to know the push was wrong.
  14. THE HYCU DATA PROTECTION TABLE Resilience for your data estate

    Vm Aw Gc Gv Af Nf N2 Np Ah Na De Xs Bq Cs Ad Bt M3 Gw Gh Gl 11 ML/AI & DBaaS Ns 9 Collaboration Fs Az Rd Dy Nd Re Pc 10 DevOps & ITSM Sf Cf Im Bx As Tr No 9 IAM & Security Bb 1 P LA TFORM 14 Infrastructure Nx 100+ WORKLOADS Ji Js Ja Jp Ci Tf Ac 45 En Entra ID 46 47 48 49 50 51 52 53 Ow Oc Ia Km Pm Wf Vp R5 Okta WIC Okta CIC AWS IAM Amazon KMS AWS Parameter Store AWS WAF Amazon VPC Amazon Route 53 +47 more workloads in the RCloud Marketplace → PREVENT · UNCOVER · RECOVER · hycu.com/marketplace
  15. HYCU backs up all five Azure DevOps services, and everything

    holding them together Boards Repos Pipelines Test Plans Artifacts • Work items: fields, relations, comments, attachments, full revision history • Git repos as full mirrors: all branches, optional LFS • YAML and classic pipeline definitions • Plans, suites, cases • Feed definitions and permissions • Queries, backlogs, boards • Pull requests • Project and code wikis • Teams and delivery plans • Build run history, releases, deployments • Runs with results and code coverage • Configurations • Variable groups, environments and checks • Package binaries across six feed types NuGet npm Maven PyPI • Agent queues Cargo • Dashboards and analytics views Universal ACROSS THE ORGANIZATION AND EVERY PROJECT Process templates and custom work item types Areas and iterations with permissions Security ACLs Agent and deployment pools Users, groups and memberships Installed extensions Service connections and service hooks Organization audit log
  16. The Git estate, complete GitHub Bitbucket GitLab Azure DevOps GA

    TODAY Jira Confluence Okta Entra ID Microsoft 365 Salesforce + 90 more
  17. CONTINU I T Y RECOVERY Data lost, corrupted, or compromised

    Data intact, but unreachable Every hour of waiting is lost work.
  18. R-Serve Your Most Critical Data, Always Accessible Always on, always

    ready Back to work, instantly Easy, self-serve access Familiar, iManage-like UI Permission-aware access Your cloud, your control Continuous, uninterrupted access to your business-critical data
  19. The Big Gap Do I have PHI in Confluence? Is

    there someone exporting something they shouldn’t? Did my identity configurations drift? What has John Smith deleted in the last 7 days?
  20. What if the answer has been sitting in your backups

    all along? 01 Historical 02 Complete 03 Independent
  21. Every SaaS backup captures a complete, time-stamped record of everything

    that happened. 01 02 03 04 Every file touched. Every permission changed. Every record modified. Every configuration adjusted.
  22. Breadth x Depth HYCU R-Cloud is the deepest context pool

    100+ workloads protected in the HYCU Marketplace, across hybrid infrastructure, cloud services, SaaS, and AI/ML. DEPTH PER APPLICATION: • Records and custom fields • Attachments and content • Permissions and sharing state • Configurations and policies • Identity objects and role state • Meta data and logs
  23. Q&A