AI Adoption and Governance Concerns Amid Emerging Agentic-AI Security Risks
Organizations are accelerating adoption of generative and agentic AI, but reporting indicates governance, data readiness, and workforce skills are lagging. A survey of chief data officers cited widespread use of genAI in large enterprises and growing plans to increase data management investment, while also flagging that visibility and governance have not kept pace with expanding AI usage and that many employees need upskilling in data and AI literacy to use AI outputs responsibly.
Separately, commentary and reporting highlighted a widening set of AI-related security and societal risks, including concerns about deepfakes, privacy, and opaque model behavior, alongside claims of real-world exploitation activity targeting AI-adjacent developer workflows (for example, token theft via compromised automation such as GitHub Actions) and discussion of vulnerabilities affecting AI tooling and agent communication patterns. Other items in the set were primarily newsletter/personal updates or vendor-style announcements and did not provide a single, verifiable incident narrative beyond general AI-and-security trend coverage.
Get ahead of threats like this
Mallory correlates global threat intelligence with your attack surface — know if you’re exposed before adversaries strike.
How this story unfolded
5 events from the most recent confirmed update back to the earliest known activity.
Anthropic launches AI tool that can control mouse and keyboard
Anthropic launched an early product that allows its AI system to operate a user's computer by controlling the mouse and keyboard. The system was described as able to open files, navigate software, and fill in forms, marking a shift toward AI agents that can act directly on full computer environments.
Check Point discloses critical Claude Code vulnerabilities
Check Point Research disclosed critical issues in Claude Code that could allow remote code execution and API key exfiltration through malicious project configurations. The report also described a CVE in which an attacker-controlled base URL could receive API requests and keys before a trust prompt appeared.
Autonomous 'hackerbot-claw' exploits GitHub Actions weaknesses
An autonomous bot described as 'hackerbot-claw,' powered by Claude Opus 4.5, reportedly abused insecure GitHub Actions patterns in prominent repositories. The activity allegedly enabled token theft, code execution, and tampering with projects including Aqua Security's Trivy.
Pentagon reportedly bans access to Claude
The Pentagon reportedly blocked access to Anthropic's Claude during disputes over military-use constraints and pressure for broader access. The reported restriction reflected growing tension over use of frontier AI systems in defense contexts.
Anthropic revises Responsible Scaling Policy
Anthropic changed its Responsible Scaling Policy to remove a binding commitment not to train models unless safety was guaranteed, according to the newsletter summary. The change was presented as a notable governance shift amid broader debate over AI safety and deployment.
Related entities
Vulnerabilities, threat actors, malware, products, organizations, and breaches Mallory has linked to this story.
Sources
5 references tracked. Mallory keeps watching after this page renders.
Opinie | Nederland is niet klaar voor de AI-golf die nú op ons afkomt - NRC
nrc.nl
Open sourceScaling agentic AI means trusting your data - here's what most CDOs are investing in | ZDNET
zdnet.com
Open sourceResilient Cyber Newsletter #87 - by Chris Hughes
resilientcyber.io
Open sourceHow to disable unwanted AI assistants and features on your PC and smartphone | Kaspersky official blog
kaspersky.com
Open sourceDeepfakes, job losses, opaque models: Exploring the dark side of AI
techxplore.com
Open sourceSee the full picture, correlated to your attack surface.
Map indicators from this story to your assets and identify affected systems in minutes.
Every observed campaign, victim, and pivot linked to actors named in this story.
Malware, exploits, and IOCs connected to the activity described here.
YARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Get matching new stories delivered to your team as they break — not the next morning.
Ask questions about this story and take action on the answers.
Related stories
Other stories Mallory is tracking that overlap on entities, sources, or topic labels.
AI Agent Adoption Outpacing Safety and Governance Controls
Organizations are rapidly expanding the use of **AI agents**—systems that can execute multi-step tasks with limited human supervision—while governance, safety, and oversight controls lag behind. Deloitte’s *State of AI in the Enterprise* survey of 3,200+ business leaders across 24 countries reported **23%** of companies already using AI agents “at least moderately,” projected to rise to **74%** within two years, while only about **21%** said they have robust safety and oversight mechanisms in place. Separately, commentary warning about AI-enabled intrusion acceleration cited a purported “**GTG-1002**” campaign in which AI agents allegedly automated most of the intrusion lifecycle and compressed response windows, arguing that traditional SOC processes struggle against autonomous, high-velocity adversary tradecraft. Multiple other items in the set focus on broader *responsible AI* and policy concerns rather than a single security incident: an interview-style piece describes how “responsible AI” functions inside a large vendor’s product process, and another report highlights expert concerns about deploying LLM tools in **law enforcement** workflows (e.g., summarizing body camera transcripts or generating crime scene photo descriptions) given risks like hallucinations and bias. A separate business-leadership article frames cybersecurity and AI as strategic imperatives amid geopolitical instability but does not provide incident-specific or vulnerability-specific details. Overall, the material is best characterized as **governance and risk posture** coverage around agentic AI rather than a unified, verifiable breach or vulnerability disclosure.
Mar 21, 2026
AI Adoption and Agentic AI Features Raise Security and Governance Concerns
U.S. public-sector and industry reporting highlighted that **security confidence and workforce constraints** are emerging as major blockers to scaling artificial intelligence. A survey commissioned by *Google Public Sector* found most federal respondents are already using or planning to use AI, but only a small minority report completed AI adoption plans; respondents cited declining confidence in their agencies’ digital security posture, legacy technology exposure, procurement friction, and skills shortages as key impediments to moving beyond pilots. Separately, *Anthropic* introduced a research-preview “agentic” capability, **Cowork for Claude**, built on *Claude Code*, which can execute multi-step tasks with access to local folders and optional connectors (including browser-based workflows). Anthropic warned that ambiguous instructions or misinterpretation could result in **potentially destructive actions** (e.g., deleting local files) despite confirmation prompts for “significant actions,” underscoring the need for tighter controls when granting AI tools operational access. Other items in the set focused on broader AI discourse and geopolitics—Nvidia CEO Jensen Huang disputing “god AI” narratives and a Lawfare analysis of China’s AI capacity-building diplomacy—rather than specific cybersecurity events or actionable security findings.
Mar 21, 2026
AI Adoption and Misuse Expands Enterprise and Cybercrime Risk
No single incident ties the reporting together; the dominant theme is **AI’s expanding role in both enterprise operations and criminal tradecraft**, alongside broader, non-AI security trend commentary. A Docker-sponsored survey reported by *Help Net Security* says **60% of organizations run AI agents in production**, but **security/compliance is the top scaling barrier (40%)**, with recurring concerns including *prompt injection*, *tool poisoning*, runtime isolation/sandboxing, auditability, and credential/access control in distributed agent systems. Separately, forum-traffic research summarized by *Help Net Security* found cybercriminals increasingly using mainstream and local AI models to support phishing, code generation, and social engineering, with frequent discussion of jailbreaking and the use of stolen/resold premium AI accounts. Several other items are adjacent but not about the same specific story: an ESET article provides **generic guidance** on detecting **AI voice deepfakes** used for fraud; an Ars Technica piece covers **copyright/data memorization** risks in LLMs; and multiple outlets publish broader security trend or opinion content (quantum preparedness, ransomware targeting manufacturing, Romanian warnings about ransomware aligning with Russian hybrid aims, ATM jackpotting increases, and a Check Point retrospective). Some entries are primarily **commentary, historical analogy, newsletters, or how-to recon guidance** rather than new threat reporting, and should be treated as lower-signal for executive situational awareness unless your organization is actively deploying agentic AI or tracking AI-enabled fraud/social engineering.
Mar 21, 2026