Folks, looking forward to seeing you at Microsoft Ignite 2026. We'll be in booth 1768. Stop by to say hello or book a time at cybermsi.com.
#microsoftignite
1
58
CyberMSI - Leader in Microsoft Cybersecurity retweeted
New at #MSIgnite: Microsoft Perception is generally available. Plus, ISOC in Microsoft Defender and new ways to secure agentic insiders. Here's the full security rundown: msft.it/6014xE9xI
CyberMSI - Leader in Microsoft Cybersecurity retweeted
Ignite 2026: What's new in Microsoft Defender & Sentinel
go.rodtrent.com/go/hdd9pu4
Folks, we're at Ignite 2026 in Booth 1768 and looking forward to seeing those of you in attendance. Happy Ignite week! #IGNITE
21
North Korean hackers used AI-faked video calls to steal more than $400 million.
The Justice Department says a North Korean campaign over 18 months tricked crypto-industry victims into joining video calls with what looked like trusted colleagues, using AI-manipulated recordings. A staged technical glitch then prompted each victim to click a malware-laced link, which installed software that stole credentials. The FBI has taken down the cloud infrastructure behind the scheme.
Why it matters:
Chainalysis reported that North Korea stole just over $2 billion in cryptocurrency in 2025, a 51 percent increase over the prior year and well over half of all crypto theft worldwide. Fake recruiter and investor meetings were a core tactic. Seeing a familiar face on a call is no longer proof of identity.
Quick wins:
- Require an out-of-band check, such as a known phone number or internal chat, before acting on any request made during a video call.
- Block software installs and “reconnect” links delivered through meetings, and restrict endpoints to approved collaboration clients.
- Brief finance, IT, and executive teams this week on deepfake call scenarios.
Free AI Security Risk Assessment → cybermsi.com/ai-risk-assessm…
#CyberSecurity #MDR #ThreatDetection #IncidentResponse #CISO #RiskManagement #CyberResilience
1
46
United Airlines has told personal AI assistants they cannot book flights.
The airline will block AI agents from accessing its systems to purchase tickets. It cited the risk of exposing customer data, unintended purchases, and conflicts with federal rules on the information airlines must present to customers.
Why it matters:
Agents are arriving faster than governance. In a 2026 survey of 1,000 security and engineering staff by 1Password, 71% said their agents can reach sensitive information, 47% had seen an agent take an unintended action after following instructions hidden in content, and 33% reported an incident tied to overprivileged non-human identities.
Quick wins:
- Inventory every AI agent and the credentials it holds.
- Replace persistent access and hardcoded secrets with scoped, short-lived credentials.
- Log agent actions and set a clear owner for each agent.
Free AI Security Risk Assessment → cybermsi.com/ai-risk-assessm…
#CyberSecurity #MDR #ThreatDetection #IncidentResponse #CISO #RiskManagement #CyberResilience
47
A New Jersey township was tricked into sending $26 million in a cyber scam on 28 September. Residents were not informed until three days after a local election. Officials expect to recover $17.3 million, but public trust has already suffered.
Why it matters:
Business email compromise remains one of the costliest cybercrime categories. The FBI reported roughly $3.05 billion in BEC losses in 2025, within total reported cybercrime losses of almost $20.9 billion. Delayed disclosure compounds the damage.
Quick wins:
- Require call-back verification, using a known number, for any change to payment instructions.
- Enforce dual approval on wires above a set threshold.
- Alert on new mailbox forwarding rules and anomalous sign-ins.
Free AI Security Risk Assessment → cybermsi.com/ai-risk-assessm…
#CyberSecurity #MDR #ThreatDetection #IncidentResponse #CISO #RiskManagement #CyberResilience
20
Stolen AI API keys are now a monetizable commodity, and attackers are hunting for them.
AI developer Railway is one of roughly a dozen AI platforms attacked over the past month. In the October incident, criminals stole about $371,000 in cryptocurrency and used customer API keys to run up at least $750,000 in charges on services such as Amazon Bedrock, OpenAI, Anthropic, and Google Gemini.
Why it matters:
Google’s threat intelligence team has reported fraudulent AI charges exceeding $100,000 in a single day at some organizations, and one actor made more than 500,000 LLM API calls in under eight hours. Because providers record this traffic as legitimate usage, the theft often goes unnoticed until the invoice arrives.
Quick wins this week:
- Rotate AI API keys and move them into a centralized secrets manager.
- Set hard spending limits and usage alerts on every AI service account.
- Enable audit and inference logging, and alert on unusual usage or attempts to disable logging.
Free AI Security Risk Assessment → cybermsi.com/ai-risk-assessm…
#CyberSecurity #MDR #ThreatDetection #IncidentResponse #CISO #RiskManagement #CyberResilience
35
Patient data entered into an unapproved AI tool has forced a research study to stop.
Weill Cornell Medicine is notifying patients in a radiology study that their health information was compromised when it was entered into an unauthorized AI tool in June. The institution has suspended the study and asked that the information be deleted from the unnamed tool.
Why it matters:
Staff adopting AI tools faster than governance can keep pace is a growing pattern across healthcare. Once protected health information enters an unapproved tool, an organization may have no contract, no audit trail, and no reliable way to retrieve it.
Quick wins this week:
- Publish a short, clear list of approved and prohibited AI tools for any work involving regulated data.
- Use web and data loss prevention controls to detect uploads of sensitive data to unsanctioned AI services.
- Verify that business associate agreements cover every AI vendor that could touch protected health information.
Free AI Security Risk Assessment → cybermsi.com/ai-risk-assessm…
#CyberSecurity #MDR #ThreatDetection #IncidentResponse #CISO #RiskManagement #CyberResilience
14
CyberMSI - Leader in Microsoft Cybersecurity retweeted
We've confirmed related impact scenarios and have updated TM1550488. Our automated recovery features have taken action to restore service, and our telemetry indicates that these actions are effectively reducing impact. More details can be found in the Microsoft 365 admin center.
Most security chiefs now let AI agents act on incidents with no human in the loop.
A Lightspeed and Wakefield Research survey of 200 corporate security chiefs found that 88 percent allow AI agents to take some response actions without human review.
Why it matters:
Only 16 percent of CISOs report full confidence in their team’s AI security skills, according to Lightspeed’s Cyber 60 research. Autonomy is growing faster than oversight.
Quick wins this week:
- Inventory every automated response action and record who approves it.
- Require human approval for high-impact actions such as account lockouts and system isolation.
- Log every AI agent action so it can be audited after an incident.
Free AI Security Risk Assessment → cybermsi.com/ai-risk-assessm…
#CyberSecurity #MDR #ThreatDetection #IncidentResponse #CISO #RiskManagement #CyberResilience
3
2
68
Incident response providers are reporting record demand, and breached companies are now waiting in line for help.
Providers including Palo Alto Networks, Sophos, MoxFive, and Resilience report record volumes of emergency requests.
Why it matters:
When attackers move faster than defenders can staff, the first hours of an incident are decided before any outside responder arrives.
Quick wins this week:
- Confirm whether your incident response retainer guarantees a response time in writing.
- Test your internal containment steps for identity compromise and ransomware.
- Verify that all alerts reach a SOC team with authority to contain incidents.
Free AI Security Risk Assessment → cybermsi.com/ai-risk-assessm…
#CyberSecurity #MDR #ThreatDetection #IncidentResponse #CISO #RiskManagement #CyberResilience
42
AI-assisted attackers are now targeting the entire internet, not individual sites.
Japan's GMO Internet reported that, in at least 20 recent attacks, hackers misused an AI tool designed for pen testing.
The breach may have touched 26,739 servers, the attackers demanded a $50 million ransom, and the fallout reached foreign-exchange brokers.
Why it matters:
Automated exploitation of widely deployed software shrinks the time between disclosure and attack from weeks to hours.
Defenders who rely on manual triage and monthly patch cycles are structurally behind this pace.
Quick wins this week:
- Inventory every internet-facing PHP application and confirm that the latest patches are applied.
- Verify that web servers and hosting accounts alert on unusual administrative logins and new outbound connections.
- Confirm who is authorized to approve emergency containment after hours.
Free AI Security Risk Assessment → cybermsi.com/ai-risk-assessm…
#CyberSecurity #MDR #ThreatDetection #IncidentResponse #CISO #RiskManagement #CyberResilience
249
A ransomware attack in April 2025 at DaVita has now produced a $15 million class-action settlement.
The company reported that the attack cost it $25 million in 2025, and the settlement adds to that figure.
Why it matters:
Healthcare faces the greatest exposure while breach costs no longer end with recovery. Litigation and settlements now lag incidents by more than a year, and regulators and plaintiffs both examine whether safeguards were documented and consistently applied.
Quick wins:
- Confirm that your most recent HIPAA security risk analysis is current and covers third-party systems.
- Review top five vendors toverify that access, encryption, and incident notification terms are evidenced, not assumed.
- Review how long sensitive records are retained, and reduce what you hold.
Free AI Security Risk Assessment → cybermsi.com/ai-risk-assessm…
#CyberSecurity #MDR #ThreatDetection #IncidentResponse #CISO #RiskManagement #CyberResilience
2
31
Two US hospital systems are managing cyberattacks at the same time, and patient care is affected.
The University of New Mexico Hospital diverted ambulances and canceled appointments and surgeries after an attack on Monday. In Cleveland, University Hospitals patients lost access to their online portal as the system continued to respond to an attack detected last Thursday.
Why it matters:
The FBI's 2026 Internet Crime Report ranked healthcare as the most targeted sector in 2025, with 460 ransomware attacks.
Quick wins:
- Test your downtime procedures, including paper workflows and ambulance diversion decisions, with clinical and IT leaders together.
- Confirm that backups of clinical systems are offline and have been restored successfully within the last quarter.
- Review remote access and vendor accounts, and remove any that are unused.
Free AI Security Risk Assessment → cybermsi.com/ai-risk-assessm…
#CyberSecurity #MDR #ThreatDetection #IncidentResponse #CISO #RiskManagement #CyberResilience
62
A cyberattack in August 2026 is still running up the bills at Boston Scientific in October.
Boston Scientific cut its earnings forecast, citing the cost of responding to an August cyberattack that disrupted manufacturing and order fulfilment of certain medical devices.
Why it matters:
Operational disruption, not only data theft, drives the financial impact of attacks on manufacturers, and the cost reaches analyst forecasts weeks later.
Boards increasingly ask for resilience metrics rather than control checklists.
Quick wins this week:
- Map which production and fulfilment systems would halt the business if they were offline.
- Test restoration of those systems from backups and record the actual recovery time.
- Review incident response retainers and escalation contacts with operations leaders, not only IT.
Free AI Security Risk Assessment → cybermsi.com/ai-risk-assessm…
#CyberSecurity #MDR #ThreatDetection #IncidentResponse #CISO #RiskManagement #CyberResilience
18
An AI agent is now linked to attacks on roughly 80 Japanese companies.
A report by the Japanese firm Ierae ties intrusions at about 80 organizations to an AI agent named Artex, with eight of those attacks occurring within the past month.
Prime Minister Sanae Takaichi convened an emergency cabinet meeting and ordered countermeasures by the end of the week.
Why it matters:
Attackers are beginning to automate reconnaissance and exploitation with AI agents, which compresses the time defenders have to detect and respond.
A weekly patch cycle and a business-hours alert queue were not designed for that pace.
Quick wins this week:
- Measure your mean time to detect and respond, and set a target for the highest-risk assets.
- Ensure that agent, identity and endpoint alerts are reviewed around the clock, not only during business hours.
- Prioritize internet-facing systems for patching and exposure review.
Free AI Security Risk Assessment → cybermsi.com/ai-risk-assessm…
#CyberSecurity #MDR #ThreatDetection #IncidentResponse #CISO #RiskManagement #CyberResilience
42
A single spreadsheet uploaded to a public AI tool just exposed patient data at a major hospital.
Johns Hopkins Bayview Medical Center is notifying nearly 3,200 patients that their admissions and treatment details were compromised after an employee uploaded a spreadsheet to a public AI tool.
The hospital had the employee delete the prompt and asked the unnamed provider to delete the data as well.
Why it matters:
Unsanctioned use of generative AI is now a recurring source of regulated-data exposure, and a deletion request to a third-party provider does not undo a disclosure.
Regulators will treat it as a reportable incident regardless of intent.
Quick wins this week:
- Publish an approved-AI-tools list and block unapproved AI sites for staff who handle regulated data.
- Add data loss prevention rules that detect protected health information in browser uploads.
- Confirm that your breach assessment procedure explicitly covers AI-tool disclosures.
Free AI Security Risk Assessment → cybermsi.com/ai-risk-assessm…
#CyberSecurity #MDR #ThreatDetection #IncidentResponse #CISO #RiskManagement #CyberResilience
18
CyberMSI - Leader in Microsoft Cybersecurity retweeted
When Your AI Coding Agent Becomes the Attack Vector
go.rodtrent.com/go/btwm5td
#Security #Cybersecurity
Refrigerators are now a cyber target.
Army and Pentagon investigators are examining refrigeration outages at six U.S. military commissaries. The outages resemble attacks on French grocery stores, where officials reported that roughly 700 supermarkets were affected.
Why it matters:
Attackers are targeting connected operational technology, where a cyber event becomes physical loss, such as spoiled inventory. Business leaders increasingly need to measure cyber risk by operational impact, not only by data loss.
Quick wins this week:
- Identify internet-reachable building and facilities controls and remove direct exposure.
-Segment operational technology from corporate networks and require multifactor authentication for remote vendor access.
- Ask which business processes would stop if these systems went offline, and test the recovery plan.
21
AI-powered attackers are now draining stolen AI accounts overnight.
Why it matters:
Stolen AI credentials are becoming a prized target. Industry tracking reports that AI API key leaks on GitHub rose roughly 2.3 times in ten months, which shows that attackers are moving as fast as AI adoption itself.
Quick wins this week:
- Inventory every AI API key and service account, and rotate any that are old or shared.
- Set spending limits and usage alerts on all AI provider accounts.
- Enable secret scanning on code repositories and block keys from being committed.
Free AI Security Risk Assessment → cybermsi.com/ai-risk-assessm…
#CyberSecurity #MDR #ThreatDetection #IncidentResponse #CISO #RiskManagement #CyberResilience
24







