Loading story…
Microsoft AI released a draft 37-page Humanist AI Code of Conduct that would govern MAI model development from 2027, with the stated aim of preserving meaningful human control over increasingly capable systems. The company is seeking public feedback for six weeks and expects to publish a revised code by year-end; the framework would inform training, technical controls, monitoring, and organizational practices.
The draft prohibits practical cyberattack enablement, including working exploit code, attack tools, intrusion and evasion procedures, and targeting or planning methods, while allowing authorized defensive work such as vulnerability discovery, malware analysis, and proof-of-concept testing. It also establishes that untrusted webpages, files, tool outputs, and AI messages cannot direct a model; requires agents to use minimum-privilege, constrained, reversible actions; bars self-escalation of access; and mandates equivalent safeguards for delegated agents. Limited cybersecurity, national-security, public-safety, and dual-use cases will receive enhanced review.
Track how attackers are adapting to this technology.
3 events from the most recent confirmed update back to the earliest known activity.
Microsoft AI published a draft Humanist AI Code of Conduct for MAI Models. The draft establishes non-overridable restrictions on cyberattack enablement, human-control requirements for autonomous agents, and a chain of command for handling untrusted external content.
Microsoft said it launched superintelligence efforts using humanist principles intended to keep humanity in control of advanced AI.
Microsoft AI said it is establishing a Humanist AI Evaluations program and developing an initial evaluation set to assess whether future MAI models conform to its Code of Conduct. The preliminary methodology defines 15 desired behaviors, testable sub-behaviors, and synthetic scenarios for scoring model responses.
Vulnerabilities, threat actors, malware, products, organizations, and breaches Mallory has linked to this story.
Follow how adversaries are adapting to this technology, and where it touches your stack today.
6 references tracked. Mallory keeps watching after this page renders.
itpro.com
Open sourcehelpnetsecurity.com
Open sourcesecurityweek.com
Open sourcefoxbusiness.com
Open sourcemicrosoft.ai
Open sourcemicrosoft.ai
Open sourceMap 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.
Other stories Mallory is tracking that overlap on entities, sources, or topic labels.
Governments, regulators, and major AI developers moved to formalize oversight of advanced AI systems as concerns grew that frontier and agentic models could accelerate cyberattacks, create liability exposure, and outpace existing governance. In the United States, **NIST** expanded work on AI security through its `AI Agent Standards Initiative`, the `Cyber AI Profile`, and the Commerce Department’s **Center for AI Standards and Innovation (CAISI)**, while Google, Microsoft, xAI, OpenAI, and Anthropic agreed to voluntary pre-release model testing with the government. The White House debated a broader executive order after reports that models such as Anthropic’s **Mythos** and OpenAI’s cyber-focused systems could rapidly identify vulnerabilities and support complex offensive operations; the final order reportedly shortened the review window and paired model access with an AI cybersecurity clearinghouse and vulnerability-sharing measures for critical infrastructure defenders. At the same time, Europe and U.S. states advanced a patchwork of AI governance rules that mixed delayed enforcement with tougher operational expectations. The EU agreed to simplify parts of the **AI Act**, delaying some high-risk obligations while banning nudification tools and preserving strong powers for the Commission’s AI Office over general-purpose AI, even as guidance on high-risk classification and logging requirements continued to emerge. States including Illinois, Utah, Maryland, and San Jose expanded governance-first programs through chief AI officers, sandboxes, incident-investigation frameworks, red teaming, and procurement controls, while insurers, auditors, and legal experts warned that **AI-washing**, weak board oversight, poor logging, and inadequate vendor controls are increasing D&O, regulatory, and tort risk. Across policy, industry, and academia, the direction of travel was toward stronger **auditing, verification, incident reporting, and lifecycle governance** rather than reliance on voluntary model promises alone.
Aug 1, 2026
A bipartisan discussion draft in the House, the **Great American Artificial Intelligence Act of 2026**, would create the first broad U.S. federal framework for regulating frontier AI developers such as OpenAI, Anthropic, Google DeepMind, and Meta. The proposal would require large model developers to publish safety and governance frameworks, issue transparency and incident reports, assess catastrophic risks, and undergo semiannual independent audits or verification. The bill defines catastrophic risks to include mass-casualty harms, major property damage, cyberattacks, CBRN weapon assistance, and loss-of-control scenarios, while also formalizing the **Center for AI Standards and Innovation** and funding it at **$100 million annually** from fiscal 2027 through 2029. The draft would also block states for **three years** from specifically regulating AI model development, a provision that drew immediate criticism from opponents who say it could suppress stronger state protections even as it preserves generally applicable state laws and post-deployment oversight. Beyond model governance, the measure would extend the **Cybersecurity Information Sharing Act** through 2035, direct DHS to support small and rural critical infrastructure operators, fund open-source software maintainers using frontier models for vulnerability discovery and patching, and establish AI testbeds across federal labs and private partners to evaluate threats including autonomous offensive cyber capabilities and risks to critical infrastructure.
Jun 29, 2026
Major technology vendors and cybersecurity researchers are rapidly integrating artificial intelligence and automation into security operations, with Microsoft unveiling a comprehensive suite of AI-powered enhancements across its Defender ecosystem. These updates include proactive features such as Predictive Shielding for automatic attack disruption, a natural language Threat Hunting Agent, and expanded integration with third-party services like AWS and Okta. Microsoft is also addressing the growing challenge of non-human digital identities and agent sprawl, while expanding Security Copilot with dozens of new agents to automate tasks for security operations, identity, and IT teams. Meanwhile, the industry is seeing a surge in AI-driven detection engineering, with new and updated rules targeting advanced threats such as Windows defense evasion, credential access, phishing, and supply chain attacks. However, the adoption of generative AI models introduces new risks, as demonstrated by research into the Chinese DeepSeek-R1 model, which was found to generate insecure code—especially when prompted with politically sensitive topics. This raises concerns about the security implications of using foreign AI models, particularly those subject to state influence or censorship. Additionally, the threat landscape is evolving with the emergence of LLM-generated malware, adaptive AI-driven malware detection, and the use of AI in both offensive and defensive cyber operations. Security teams are urged to remain vigilant as AI technologies reshape both the tools available to defenders and the tactics employed by adversaries.
Jun 29, 2026