AI Governance Brief — 2026-09-22
Top Stories
1. Audit Committees Elevate AI Governance as a Board-Level Priority
- Source: Deloitte / Center for Audit Quality · September 22, 2026
- Summary: Deloitte and the Center for Audit Quality’s latest Audit Committee Practices Report finds that 75% of respondents now identify AI governance as a top-three priority, up from 35% a year earlier. However, only 56% say they are confident in their committees’ ability to effectively oversee AI governance.
- Why It Matters: AI governance is moving from technology and compliance functions into formal board and audit-committee oversight. The capability gap suggests that governance structures may be expanding faster than the expertise and visibility required to operate them effectively.
- URL: https://www.prnewswire.com/news-releases/enterprise-risk-management-takes-the-lead-ai-governance-gains-urgency-for-audit-committees-deloitte-center-for-audit-quality-survey-302885335.html
2. Regnology Study Finds a Large Gap Between AI Pilots and Production Regulatory Reporting
- Source: FF News · September 22, 2026
- Summary: A Regnology and Oliver Wyman study reports that 87% of financial institutions are exploring or piloting AI, while only 16% have moved these technologies into production. The research highlights traceability, reconstruction of outputs, data quality, and human oversight as critical requirements for agentic AI in regulatory reporting.
- Why It Matters: In regulated financial workflows, governance is becoming a production-enablement layer rather than simply a compliance checkpoint. The ability to prove how an AI system reached an output may become as important as the automation itself.
- URL: https://ffnews.com/news/new-research-89-of-financial-institutions-have-yet-to-embed-ai-into-regulatory-r-0f6173ad
3. Akamai Calls for Behavioral Governance of Autonomous AI Agents
- Source: GlobeNewswire / Akamai · September 22, 2026
- Summary: Akamai’s latest State of the Internet security report argues that enterprise AI risk is shifting from traditional human identity management toward governing autonomous, nonhuman entities. The report highlights gaps around AI agents, Model Context Protocol (MCP) exposure, browser-based AI tools, and sensitive information entering chatbot interactions.
- Why It Matters: Agentic AI changes governance from controlling what users can access to controlling what autonomous systems can do. Identity, permissions, runtime behavior, tool access and API actions increasingly need to be governed as a single control surface.
- URL: https://www.globenewswire.com/news-release/2026/09/22/3366103/0/en/akamai-report-securing-agentic-ai-requires-shift-to-behavioral-governance.html
4. DeepSeek Set to Brief UN Security Council on AI Risks
- Source: Reuters · September 22, 2026
- Summary: DeepSeek is expected to participate in a UN Security Council discussion on AI and international security alongside representatives from OpenAI and Anthropic. The meeting comes amid renewed discussion about coordinated safeguards for increasingly capable AI systems and further US-China talks on AI safety.
- Why It Matters: AI governance is increasingly becoming an international-security issue rather than solely a technology-policy question. The participation of both US and Chinese AI companies underscores the growing need for cross-border mechanisms covering advanced-model risks.
- URL: https://www.reuters.com/world/asia-pacific/deepseek-brief-un-security-council-ai-this-week-sources-say-2026-09-22/
5. Australia Faces Pressure to Reconsider AI Training Copyright Rules
- Source: Reuters · September 22, 2026
- Summary: OpenAI and Anthropic have urged Australia to consider a conditional approval framework for AI model training using Australian creative content. Australia has ruled out a broad copyright exemption and plans broader AI and data-centre rules from next year.
- Why It Matters: The dispute illustrates the emerging governance trade-off between AI development, creator rights, data sovereignty and investment incentives. Conditional approval models could become a template for jurisdictions seeking to balance AI infrastructure investment with stronger rights protections.
- URL: https://www.reuters.com/legal/litigation/anthropic-openai-call-australia-relax-ban-training-ai-models-2026-09-22/
6. Singapore Links Public-Sector AI Adoption to Trust, Accountability and Inclusion
- Source: OpenGov Asia · September 22, 2026
- Summary: Singapore outlined an approach to public-sector AI that emphasizes public value, trust and inclusion alongside deployment. The government is exploring AI agents to redesign workflows while maintaining accountability for important decisions, with safeguards positioned as an enabler of sustainable adoption.
- Why It Matters: Singapore’s approach reinforces a governance model in which AI deployment and accountability advance together. For enterprises, the implication is that successful agentic-AI programs increasingly require explicit ownership and controls embedded directly into operating workflows.
- URL: https://opengovasia.com/singapore-outlines-public-sector-approach-to-inclusive-ai/
7. Global AI Governance Body Proposed as Risks Outgrow National Rules
- Source: Human Resources Director · September 22, 2026
- Summary: University of Queensland professor Stan Karanasios called for an international AI governance body modeled on institutions such as the International Civil Aviation Organization and International Atomic Energy Agency. The proposal argues that fragmented national rules covering issues such as deepfakes, bias and cyber risks may be insufficient for increasingly capable AI.
- Why It Matters: The proposal reflects a broader shift toward international governance mechanisms, including common standards, independent audits and incident transparency. Whether such an institution emerges remains uncertain, but pressure for cross-border AI oversight is becoming more explicit.
- URL: https://www.hcamag.com/au/news/general/global-ai-governance-body-needed-to-address-growing-risks-expert-says/590618
8. Fastly Expands Into Runtime AI Governance and Agent API Controls
- Source: Futurum Group · September 22, 2026
- Summary: Fastly’s newly announced AI Runtime Control, AI Firewall and enhanced API Security capabilities are positioned as a unified control layer for enterprise AI traffic. The capabilities include model-call routing, token-spend visibility, rate and budget controls, prompt-injection protection and enforcement of API contracts for agentic traffic.
- Why It Matters: AI governance is increasingly moving into the runtime layer rather than remaining confined to policies and pre-deployment reviews. The convergence of security, identity, API governance and AI cost controls points toward an emerging enterprise AI control plane.
- URL: https://futurumgroup.com/insights/fastly-bets-the-edge-on-ai-governance-as-machine-traffic-tops-50/
9. Enterprise AI Governance Is Shifting Toward Operational Control
- Source: Gartner Security & Risk Management Summit / LogicGate · September 22, 2026
- Summary: AI governance discussions at the Gartner Security & Risk Management Summit are increasingly focused on operational controls for enterprise AI. LogicGate highlighted AI governance agents designed to expand review coverage and help organizations manage rapidly changing regulatory requirements.
- Why It Matters: Governance tooling is evolving from static policy documentation toward continuous monitoring, workflow automation and evidence generation. This points toward AI governance becoming an operational system rather than a periodic compliance exercise.
- URL: https://www.logicgate.ai/events/
10. Agentic AI Governance Moves Into Legal-Function Workflows
- Source: Singapore Corporate Counsel Association · September 22, 2026
- Summary: The Singapore Corporate Counsel Association’s agentic-AI program is testing practical legal workflows with human oversight, evaluation harnesses, approved-content retrieval and policy-aligned agent orchestration. The program focuses on moving from proof of concept toward a governed operating model for legal teams.
- Why It Matters: The example shows how governance requirements are becoming embedded in agent design itself: approved data sources, human-in-the-loop controls, evaluation, documentation and policy-aligned tool use. This is a practical model for scaling agentic AI in other regulated functions.
- URL: https://scca.glueup.com/event/scca-leah-agentic-ai-for-in-house-departments-186137/