Managed Detection & Response
Detection usually fails where telemetry is missing, not where analysts are slow. MDR pays off when coverage of the blind spots (edge devices, hypervisors, identity systems and now AI platforms), log retention and the provider's authority to contain are agreed before the first alert.
Part of Security, Assurance & Governance, in the Run mode. 8 sources, checked .
- global median attacker dwell time in Mandiant investigations in 2025, up from 11 days in 2024
- 14 days
- global median attacker dwell time in Mandiant investigations in 2025, up from 11 days in 2024
- Mandiant, Google Cloud, 2026(opens in a new tab)
- of intrusions Mandiant investigated in 2025 were first flagged to the victim by an outside party
- 34%
- of intrusions Mandiant investigated in 2025 were first flagged to the victim by an outside party
- Mandiant, Google Cloud, 2026(opens in a new tab)
- lower average breach cost for organizations using AI and automation extensively in security, versus none (2026 study)
- $1.93M
- lower average breach cost for organizations using AI and automation extensively in security, versus none (2026 study)
- IBM, 2026(opens in a new tab)
The problem
Attackers are fast, and some stay a long time. In Mandiant's 2025 investigations, global median dwell time rose to 14 days from 11, driven by espionage and North Korean IT worker intrusions, which had a median of 122 days.1 Organizations found the intrusion themselves 52% of the time; 34% of the time an outside party told them and 14% of the time the attacker did.1 The median time from initial access to hand-off to a second threat group fell from more than 8 hours in 2022 to 22 seconds in 2025.1
The gaps are in telemetry. Mandiant reported that cases involving the BRICKSTORM backdoor averaged 393 days of dwell time, so standard 90-day log retention left organizations unable to identify how attackers got in. Those actors targeted edge appliances and authentication protocols that endpoint agents do not reach.1 Exploits were the most common way in, at 32% of intrusions.1 Verizon's 2026 DBIR found the same shift: vulnerability exploitation, at 31% of breaches, has overtaken stolen credentials as the top point of entry for the first time in 19 years.2
AI platforms add a new surface to watch. In IBM's 2026 Cost of a Data Breach research, more than one in five organizations reported an AI-related breach, an attack on AI models or applications, and 92% of those lacked proper AI access controls.4 The global average cost of a breach reached a record $4.99 million, while organizations using AI and automation extensively in security saved $1.93 million on average compared with those using none.3
Disclosure rules make detection a board issue. US public companies must disclose a material cybersecurity incident on Form 8-K, generally within four business days of determining it is material, and must describe the board's oversight of cyber risk.5 Joint guidance from Australia's ACSC, CISA, the FBI and NSA sets a baseline for event logging to detect malicious activity,6 and NIST SP 800-61 Rev. 3 builds incident response into risk management under CSF 2.0.7
What you get
- 01
Around-the-clock monitoring and triage
Continuous monitoring of your AI and data platforms, cloud control planes and identity systems, with alerts investigated by analysts and escalated within agreed times.
- 02
AI-specific detections
Detection content for attacks on AI systems, such as prompt injection, model and data exfiltration, abuse of model API keys and unusual agent tool use, mapped to MITRE ATT&CK and MITRE ATLAS techniques.
- 03
Telemetry and retention design
A logging plan that covers edge devices, hypervisors, identity providers, model gateways and vector stores, with retention long enough to scope intrusions that run for months.
- 04
Response playbooks with pre-agreed actions
Playbooks that state which containment steps our analysts may take on their own, such as revoking a key or isolating a host, and which need your approval, plus support through investigation and recovery.
- 05
Incident facts for disclosure decisions
Timelines, scope and impact findings packaged for your legal and disclosure team, so materiality decisions under SEC and sector rules rest on evidence.
Where it goes wrong
The failure modes we design against from the first week, and the control for each.
Endpoint-only coverage
Endpoint agents do not see edge appliances, hypervisors or authentication protocols, which is where long-dwell actors operate.1 Define the telemetry scope, including AI platforms, in the contract before onboarding, and test that each source actually arrives.
Log retention shorter than dwell time
Some intrusions run for more than a year, and 90-day retention leaves you unable to find the entry point.1 Forward the logs you need for scoping to centralized, long-term storage, especially edge, hypervisor and authentication logs.
No authority to act out of hours
If the provider can only open tickets, containment waits for business hours while access can change hands between criminal groups in seconds.1 Agree in writing which actions the MDR team may take alone and who approves the rest.
AI platforms left out of scope
Model endpoints, API keys and AI data stores are often run by a separate team and missing from security monitoring. Among organizations with an AI-related breach, 92% lacked proper AI access controls.4 Bring AI systems into the asset inventory and the detection scope.
Standards and rules
SEC Form 8-K Item 1.05 and Regulation S-K Item 106
Primary text on sec.gov(opens in a new tab)Public companies must disclose material cybersecurity incidents, generally within four business days of determining materiality, and describe board oversight of cyber risk, so detection must feed disclosure quickly.
NIST SP 800-61 Rev. 3 (April 2025)
Primary text on csrc.nist.gov(opens in a new tab)Builds incident response into cybersecurity risk management as a CSF 2.0 Community Profile, superseding the 2012 Rev. 2.
Best Practices for Event Logging and Threat Detection (ACSC, CISA, FBI, NSA)
Primary text on cisa.gov(opens in a new tab)Defines a baseline for event logging so organizations can detect malicious activity, behavioral anomalies and compromised networks, devices or accounts.
The shared knowledge base of adversary tactics and techniques used to map, test and report detection coverage.
How to measure it
- Mean time to detect and mean time to contain, by severity
- Share of incidents detected internally versus reported by outside parties or attackers
- Coverage of in-scope log sources, including identity, edge, hypervisor and AI platforms
- ATT&CK technique coverage validated by testing, not just mapped on paper
- Time from detection to an incident summary your disclosure team can act on
Questions buyers ask
What is the difference between MDR and a SOC?
A security operations center is a function; MDR is a service that runs detection and response for you. A traditional managed security service often forwards alerts for your team to handle. An MDR provider investigates alerts, hunts for threats and takes agreed containment actions itself. Many organizations keep a small internal team for context and decisions and use MDR for round-the-clock coverage.
Do AI systems need different security monitoring?
They need additional coverage. AI platforms bring new assets to watch: model endpoints, API keys, prompts and outputs, retrieval data stores and agents that call tools. Attacks such as prompt injection or data theft through a model look different from malware. Add these sources to your logging plan and write detections for AI-specific techniques, alongside the usual identity, cloud and endpoint monitoring.
How does MDR help with the SEC four-day incident disclosure rule?
The four business days generally start when you determine an incident is material, not when it begins, but you cannot make that call without facts. A good MDR provider gives you a timeline, scope and impact assessment quickly, so your disclosure committee decides on evidence. Agree in advance how the provider's findings reach legal counsel and who owns the materiality decision.
Sources
Every figure on this page comes from the sources below. We checked each link on .
- 1.M-Trends 2026 Executive Edition(opens in a new tab)Mandiant, Google Cloud, March 2026 · services.google.com
- 2.Vulnerability exploitation top breach entry point, 2026 industry-wide DBIR finds(opens in a new tab)Verizon, May 19, 2026 · verizon.com
- 3.Cost of a Data Breach Report 2026(opens in a new tab)IBM, July 2026 · ibm.com
- 4.AI-powered adversaries and the enterprise risk challenge: Preparing for the new reality(opens in a new tab)IBM, July 29, 2026 · ibm.com
- 5.SEC Adopts Rules on Cybersecurity Risk Management, Strategy, Governance, and Incident Disclosure by Public Companies(opens in a new tab)U.S. Securities and Exchange Commission, July 26, 2023 · sec.gov
- 6.Best Practices for Event Logging and Threat Detection(opens in a new tab)Cybersecurity and Infrastructure Security Agency, with ASD's ACSC, FBI and NSA, August 21, 2024 · cisa.gov
- 7.SP 800-61 Rev. 3, Incident Response Recommendations and Considerations for Cybersecurity Risk Management: A CSF 2.0 Community Profile(opens in a new tab)National Institute of Standards and Technology, April 2025 · csrc.nist.gov
- 8.MITRE ATT&CK(opens in a new tab)The MITRE Corporation · attack.mitre.org