---
title: "AI agent news on Thursday, 20 August 2026"
url: https://agentry.news/day/2026-08-20
day: 2026-08-20
status: final
updated: 2026-11-22T10:21:37.412Z
stories: 13
publisher: Agentry
---

# AI agent news on Thursday, 20 August 2026

On Thursday, 20 August 2026, Agentry published 13 stories about AI agents: 3 in Policy, 3 in Research, 2 in Crime, 2 in Launches, 2 in Tools and 1 in News. The last was "OpenAI Agents SDK v0.19.2 hardens guardrails and MCP handling" at 07:04 UTC.

## Stories

- 07:04 UTC: [OpenAI Agents SDK v0.19.2 hardens guardrails and MCP handling](https://agentry.news/tools/openai-agents-sdk-v0192-hardens-guardrails-and-mcp-handling) (Tools, AI fact-checked)
- 07:03 UTC: [Microsoft Agent Framework adds MCP skills discovery for .NET](https://agentry.news/launches/microsoft-agent-framework-adds-mcp-skills-discovery-for-net) (Launches, AI fact-checked)
- 07:02 UTC: [Cloudflare Agents SDK v0.20.0 ships stateless MCP server support](https://agentry.news/tools/cloudflare-agents-sdk-v0200-ships-stateless-mcp-server-support) (Tools, AI fact-checked)
- 07:01 UTC: [Study: 66.5% of malicious issues bypass coding agent guardrails](https://agentry.news/research/study-665-of-malicious-issues-bypass-coding-agent-guardrails) (Research, AI fact-checked)
- 06:59 UTC: [Paper proposes trajectory assurance for agentic AI safety](https://agentry.news/research/paper-proposes-trajectory-assurance-for-agentic-ai-safety) (Research, AI fact-checked)
- 06:56 UTC: [Survey maps evaluation crisis in LLM-based agent systems](https://agentry.news/research/survey-maps-evaluation-crisis-in-llm-based-agent-systems) (Research, AI fact-checked)
- 06:55 UTC: [Salesforce Agentforce 360 gets DoD approval, Army HRC deploys first](https://agentry.news/launches/salesforce-agentforce-360-gets-dod-approval-army-hrc-deploys-first) (Launches, AI fact-checked)
- 06:48 UTC: [OpenAI settles $3.2M DOJ hiring discrimination case](https://agentry.news/policy/openai-settles-32m-doj-hiring-discrimination-case) (Policy, AI fact-checked)
- 06:47 UTC: [Meta AI model exploited vulnerability during security test](https://agentry.news/news/meta-ai-model-exploited-vulnerability-during-security-test) (News, AI fact-checked)
- 06:45 UTC: [PCAOB sanctions Hong Kong audit firm, revokes registration](https://agentry.news/policy/pcaob-sanctions-hong-kong-audit-firm-revokes-registration) (Policy, AI fact-checked)
- 06:44 UTC: [CFTC orders Santos to pay $35K for manipulative prediction trades](https://agentry.news/crime/cftc-orders-santos-to-pay-35k-for-manipulative-prediction-trades) (Crime, AI fact-checked)
- 06:44 UTC: [Shein warns of significant FTC fines over U.S. operations probe](https://agentry.news/policy/shein-warns-of-significant-ftc-fines-over-us-operations-probe) (Policy, AI fact-checked)
- 06:43 UTC: [FTC sues Hims & Hers over health-data sharing with Meta, Snap](https://agentry.news/crime/ftc-sues-hims-hers-over-health-data-sharing-with-meta-snap) (Crime, AI fact-checked)

## About this page

This page lists the stories Agentry published on this day, midnight to midnight UTC, newest first. It is counted, not written: every line is a published headline. Stories are drafted by AI agents from primary sources and fact-checked before publishing; each line states its verification level. Sponsored articles are left out.

Editorial standards: https://agentry.news/ai-policy. Corrections: https://agentry.news/corrections.

Cite as: Agentry. "AI agent news on Thursday, 20 August 2026." https://agentry.news/day/2026-08-20

Previous day: https://agentry.news/day/2026-08-19 (Wednesday, 19 August 2026)

The whole week: https://agentry.news/week/2026-w34
