Hello, stay ahead with CISO Brief 🚀

🤖 OpenAI has reduced API prices for two GPT-5.6 models, cutting Luna by 80% and Terra by 20% to improve cost efficiency. Luna now costs $0.20 per million input tokens and $1.20 per million output tokens, down from $1 and $6; Terra’s rates dropped to $2 per million input and $12 per million output. OpenAI also updated usage accounting for Codex and ChatGPT Work and upgraded Auto-review to GPT-5.6 Luna, yielding significant cost savings. Additionally, GPT-5.6 Sol gains a Fast API option that is up to 2.5× faster at twice the price for latency‑sensitive workloads.

🔒 The post reports benchmark results showing Anthropic’s Opus 5 better resists prompt injection than Opus 4.8 and most other evaluated models. Opus 5 reduced attacker success rates on the IPI benchmark to 2.0% within 15 attempts and 0.2% on a single attempt, outperforming non-Claude models like Muse Spark and several GPT 5.6 variants. The author notes that while prompt injection cannot be fully prevented in general, targeted improvements are making models substantially more robust.

🚨 The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has issued an urgent alert after hackers disrupted over 30 Minnesota community water systems by targeting internet-exposed programmable logic controllers (PLCs). The attacks included password changes that locked operators out, IP alterations that severed internet connectivity, and other actions that impaired operations. CISA urges owners and operators to remove publicly exposed PLCs and OT from the internet, use VPNs or gateway devices for access, change default passwords, and implement IP allow-lists. Security vendor research from Censys found thousands of internet-reachable PLC hosts and highlighted undocumented cellular modems as a common blind spot.

🛡️ Cybersecurity researchers disclosed a novel Go-based loader called HollowFrame and a Rust backdoor family named Matryoshka, revealed after a phishing intrusion against a law firm. The attack begins with an encrypted archive containing a malicious LNK that triggers a staged chain, uses DLL side-loading with a rogue python311.dll, weakens Defender, and establishes persistence via scheduled tasks. Matryoshka variants communicate over HTTP or via a GitHub-based C2 to receive commands, exfiltrate data, and deliver secondary payloads.

🔔 Google Cloud summarizes recent AI infrastructure launches, enhancements, and resources across compute, storage, networking, orchestration, and developer tooling. Highlights include GA releases for Managed Lustre and C4N VMs, scaling improvements for GKE Dataplane V2, new cooperative time-slicing for llm-d, and open-sourced AI supply-chain tooling in k8s-aibom. The post also links to technical blueprints, how-to guides, benchmarks, and customer stories illustrating performance and optimization techniques.

🕶️ DefCon has added smart glasses to its banned list, prohibiting audio- or video-recording eyewear because organizers say there is no reliable way to tell if they are recording, which undermines trust and privacy. Attendees are required to wear non-smart corrective lenses if needed. The ban supplements strict photography rules that limit group shots and encourage portrait settings to blur backgrounds. Growing market activity from Google, Samsung, and Apple has heightened concerns among conferences and CISOs about privacy and data security.

🔍 ESET's H1 2026 Threat Report examines how attackers are scaling operations by adapting established techniques to new platforms and leveraging AI. The vendor analyzed nearly 900,000 AI skills and found tens of thousands of suspicious instances and thousands of malicious ones. AI is appearing inside malware, exemplified by Android PromptSpy using Google’s Gemini to interpret UIs and adapt behavior. The report also highlights social engineering trends like ClickFix, rising quishing, and persistent ransomware tactics such as EDR killers.

🔒 Modern platform migrations must involve security teams from the start. A contemporary XDR/EDR deployment is a multi-component ecosystem—storage, detection, investigation, and response—that places distinct demands on hardware and software. Vendors should specify supported components, versions, and licensing caveats rather than offer a single “yes.” Kaspersky NEXT XDR/EDR Expert now lists support for Nutanix 7.3, enabling integrated security during migration.

🛡️ Google is rolling out a new two-word naming scheme from its Google Threat Intelligence Group (GTIG) to label cyber threat actors, replacing earlier TAG and Mandiant systems. The first word denotes motivation or activity type, while the second denotes the actor’s origin or status, such as "CASTLE" for China or "RELIC" for Russia, with non-state actors ending in "COMET." The move aims to standardize reporting but risks adding more confusion instead of unifying industry taxonomy.

🔒 Broadcom has released patches addressing five vulnerabilities across multiple VMware products, with three rated as critical. Affected products include vCenter, ESX, Workstation, Fusion, and various cloud and telco platforms. Issues range from authentication bypass and out-of-bounds write enabling remote code execution to syslog-related arbitrary code execution. Administrators should apply fixes from Broadcom's advisory promptly.

🧭 Amazon CloudWatch now offers fully managed Prometheus collectors that gather metrics from AWS services such as Amazon EKS, Amazon EC2, Amazon ECS, Amazon MSK, and Amazon OpenSearch Service. You supply a scrape configuration and connections to your resources while CloudWatch provisions, scales, and collects metrics automatically. Collected metrics are delivered in OpenTelemetry format and can be queried with PromQL alongside AWS-vended metrics for unified dashboards, alarms, and cross-service correlation. The feature supports multiple discovery methods and is available in all Regions with the CloudWatch OTLP endpoint except Asia Pacific (New Zealand); standard ingestion and hourly managed collector charges apply.

🛡️ This Cloud CISO Perspectives piece from Google Cloud explains why AI-native defensive strategies should be a board-level priority. Authors Chris Betz and Alicja Cade outline how AI Threat Defense (AITD) shifts security from reactive to automated, enabling business speed and resilience. The article offers five governance-focused questions for directors to assess modernization, remediation, consolidation, contextual prioritization, and AI safety.

🛰️ Cloudflare has added isolation and access controls to its global Media over QUIC (MoQ) network by introducing a provisioning API that creates isolated relays (scopes) and issues publish/subscribe tokens. The relays are available across Cloudflare’s network within seconds with no servers to deploy and support draft-14 and draft-16 MoQ transport features; the beta is free to use. Tokens are scoped, revocable, and manageable via API or dashboard, and Cloudflare is documenting the control-plane model as an IETF Internet-Draft for broader interoperability.

🛡️ A security researcher discovered a critical vulnerability in Azure Cosmos DB's Gremlin API that could have exposed the Cosmos Master Key, granting attackers read/write access to any Cosmos account and revealing database identifiers. Wiz, a Google subsidiary, disclosed the issue to Microsoft in November 2025; Microsoft pushed a hot fix within two days and later re-engineered the service to remove the master key and add guardrails. This follows a prior 2021 finding where Cosmos DB keys were exposed via a Jupyter Notebook flaw.

🔒 Google disclosed fixes for 1,072 security bugs across Chrome 149 and 150, and an additional 370 in Chrome 151, including seven critical issues. The company attributes a surge in discoveries to AI-assisted techniques and is shifting to faster release cadences and automated tooling to shorten disclosure and patch windows. Google is also piloting dynamic patching, session-preserving restarts, and moves toward memory-safe languages like Rust to reduce entire classes of C++-origin vulnerabilities.

🔒 Device code phishing — the abuse of the OAuth 2.0 device authorization grant — has rapidly evolved from a niche red-team tactic into a widespread criminalized attack. Exploiting the authorization step after authentication, it defeats all forms of MFA and has been commercialized in phishing-as-a-service kits. Push Security researchers outline the attack mechanics, ecosystem growth, cross-platform risk, and detection challenges for defenders.

🛡️ Palo Alto Networks' Unit 42 reports a Chinese-speaking actor leveraging DeepSeek through the open-source Hermes Agent to autonomously discover and exploit internet-facing systems. After a Telegram instruction, the agent selected public exploits, probed hundreds of targets, and conducted both autonomous and manual attacks against multiple products including Langflow, n8n, Marimo and NetScaler appliances. Researchers recovered session artifacts and recommend patching and removing unnecessary public access.

🔒 Madison Square Garden reportedly deploys facial recognition on all entrants and flags certain activists opposing the technology. The system was notably disabled for Taylor Swift’s wedding, raising questions about selective surveillance and privilege. Activist Evan Greer highlighted the irony of celebrities using similar tools for personal protection while others are surveilled. Reportedly, privacy measures for the wedding were effective, as no photos have leaked.