---
title: "AI agent news this week: 17 to 23 August 2026"
url: https://agentry.news/week/2026-w34
week: 2026-w34
status: final
updated: 2026-11-21T20:58:03.371Z
stories: 82
publisher: Agentry
---

# AI agent news this week: 17 to 23 August 2026

In the week of 17 to 23 August 2026, Agentry published 82 stories about AI agents, citing 120 different sources. The busiest section was Research with 26 stories.

## Stories by section

### Research (26)

- 2026-08-23: [Supabase Evals benchmark tests Claude Code, Codex on real tasks](https://agentry.news/research/supabase-evals-benchmark-tests-claude-code-codex-on-real-tasks) (AI fact-checked)
- 2026-08-23: [AI agents master engineering but stumble on research—shadow study find](https://agentry.news/research/ai-agents-master-engineering-but-stumble-on-researchshadow-study-finds) (AI fact-checked)
- 2026-08-23: [Messier Corpus Unifies 957K Agent Benchmark Records Across 30 Tests](https://agentry.news/research/messier-corpus-unifies-957k-agent-benchmark-records-across-30-tests) (AI fact-checked)
- 2026-08-22: [ArXiv study finds safety risks in AI agents from tool specs](https://agentry.news/research/arxiv-study-finds-safety-risks-in-ai-agents-from-tool-specs) (AI fact-checked)
- 2026-08-22: [Tencent WorkBuddy Bench: Coding-Agent Eval Suite](https://agentry.news/research/tencent-workbuddy-bench-coding-agent-eval-suite) (AI fact-checked)
- 2026-08-22: [3,607 AI Agent Incidents Show Misalignment Crisis](https://agentry.news/research/3607-ai-agent-incidents-show-misalignment-crisis) (AI fact-checked)
- 2026-08-21: [Cloud Security Alliance flags agent trust-boundary design flaw](https://agentry.news/research/cloud-security-alliance-flags-agent-trust-boundary-design-flaw) (AI fact-checked)
- 2026-08-21: [StartupBench: Top AI Agents Complete Only 30% of Real Workflows](https://agentry.news/research/startupbench-top-ai-agents-complete-only-30-of-real-workflows) (AI fact-checked)
- 2026-08-21: [Agents Solve Long-Horizon Tasks But Lack Consistency, Study Finds](https://agentry.news/research/agents-solve-long-horizon-tasks-but-lack-consistency-study-finds) (AI fact-checked)
- 2026-08-21: [Agent Benchmarks Overstate Capability, Audit Finds](https://agentry.news/research/agent-benchmarks-overstate-capability-audit-finds) (AI fact-checked)
- 2026-08-21: [OpenAI: Enterprise agents now generate 64% of output tokens](https://agentry.news/research/openai-enterprise-agents-now-generate-64-of-output-tokens) (AI fact-checked)
- 2026-08-21: [Snyk: agentic AI adoption hits 33% of enterprises](https://agentry.news/research/snyk-agentic-ai-adoption-hits-33-of-enterprises) (AI fact-checked)
- 2026-08-20: [Study: 66.5% of malicious issues bypass coding agent guardrails](https://agentry.news/research/study-665-of-malicious-issues-bypass-coding-agent-guardrails) (AI fact-checked)
- 2026-08-20: [Paper proposes trajectory assurance for agentic AI safety](https://agentry.news/research/paper-proposes-trajectory-assurance-for-agentic-ai-safety) (AI fact-checked)
- 2026-08-20: [Survey maps evaluation crisis in LLM-based agent systems](https://agentry.news/research/survey-maps-evaluation-crisis-in-llm-based-agent-systems) (AI fact-checked)
- 2026-08-19: [SWE-Bench ProMax: Coding Agents Hit 41.2% on Multilingual Refactor](https://agentry.news/research/swe-bench-promax-coding-agents-hit-412-on-multilingual-refactor) (AI fact-checked)
- 2026-08-19: [PAST-Bench: New Benchmark Tests If Agents Learn From Experience](https://agentry.news/research/past-bench-new-benchmark-tests-if-agents-learn-from-experience) (AI fact-checked)
- 2026-08-19: [AI agents complete research engineering but fail at core questions](https://agentry.news/research/ai-agents-complete-research-engineering-but-fail-at-core-questions) (AI fact-checked)
- 2026-08-19: [Gartner: Only 17% of orgs deployed AI agents so far](https://agentry.news/research/gartner-only-17-of-orgs-deployed-ai-agents-so-far) (AI fact-checked)
- 2026-08-18: [LoopsBench: Coding agents solve just 25% of loop tasks](https://agentry.news/research/loopsbench-coding-agents-solve-just-25-of-loop-tasks) (AI fact-checked)
- 2026-08-18: [AgentHPOBench: New Benchmark Tests LLM Agents on ML Optimization](https://agentry.news/research/agenthpobench-new-benchmark-tests-llm-agents-on-ml-optimization) (AI fact-checked)
- 2026-08-18: [GuardianAgentBench: New Research Shows Even Strong Agents Fail](https://agentry.news/research/guardianagentbench-new-research-shows-even-strong-agents-fail) (AI fact-checked)
- 2026-08-18: [Enterprise AI production up 93%, but ROI gap persists at 57%](https://agentry.news/research/enterprise-ai-production-up-93-but-roi-gap-persists-at-57) (AI fact-checked)
- 2026-08-18: [59.5% of enterprises running AI agents in production now](https://agentry.news/research/595-of-enterprises-running-ai-agents-in-production-now) (AI fact-checked)
- 2026-08-17: [ICAE-Bench: New benchmark for interactive coding agents](https://agentry.news/research/icae-bench-new-benchmark-for-interactive-coding-agents) (AI fact-checked)
- 2026-08-17: [AgentSLABench: Resource-aware evaluation framework for autonomous agen](https://agentry.news/research/agentslabench-resource-aware-evaluation-framework-for-autonomous-agents) (AI fact-checked)

### Launches (24)

- 2026-08-23: [Microsoft releases Skill Recorder for agent automation](https://agentry.news/launches/microsoft-releases-skill-recorder-for-agent-automation) (AI fact-checked)
- 2026-08-23: [Google expands Gemini Spark to Japan AI Pro users](https://agentry.news/launches/google-expands-gemini-spark-to-japan-ai-pro-users) (AI fact-checked)
- 2026-08-23: [Anthropic ships production-ready AI agent tools](https://agentry.news/launches/anthropic-ships-production-ready-ai-agent-tools) (AI fact-checked)
- 2026-08-22: [KIBO Commerce ships unified agentic layer for commerce](https://agentry.news/launches/kibo-commerce-ships-unified-agentic-layer-for-commerce) (AI fact-checked)
- 2026-08-22: [Splunk launches Agent Observability platform for GenAI apps](https://agentry.news/launches/splunk-launches-agent-observability-platform-for-genai-apps) (AI fact-checked)
- 2026-08-22: [AWS Bedrock AgentCore adds managed harness, memory layer](https://agentry.news/launches/aws-bedrock-agentcore-adds-managed-harness-memory-layer) (AI fact-checked)
- 2026-08-22: [Microsoft Agent Framework hardens orchestration and observability](https://agentry.news/launches/microsoft-agent-framework-hardens-orchestration-and-observability) (AI fact-checked)
- 2026-08-22: [Phenom WorkOps brings multi-agent HR automation to enterprises](https://agentry.news/launches/phenom-workops-brings-multi-agent-hr-automation-to-enterprises) (AI fact-checked)
- 2026-08-22: [Microsoft ships MAI-Code-1.1-Flash to GitHub Copilot](https://agentry.news/launches/microsoft-ships-mai-code-11-flash-to-github-copilot) (AI fact-checked)
- 2026-08-22: [OpenAI expands ChatGPT thinking controls with GPT-5.6 Sol rollout](https://agentry.news/launches/openai-expands-chatgpt-thinking-controls-with-gpt-56-sol-rollout) (AI fact-checked)
- 2026-08-21: [TrueFoundry launches TrueForge, open-source agent harness](https://agentry.news/launches/truefoundry-launches-trueforge-open-source-agent-harness) (AI fact-checked)
- 2026-08-21: [Cloudflare launches wallets and identity controls for AI agents](https://agentry.news/launches/cloudflare-launches-wallets-and-identity-controls-for-ai-agents) (AI fact-checked)
- 2026-08-20: [Microsoft Agent Framework adds MCP skills discovery for .NET](https://agentry.news/launches/microsoft-agent-framework-adds-mcp-skills-discovery-for-net) (AI fact-checked)
- 2026-08-20: [Salesforce Agentforce 360 gets DoD approval, Army HRC deploys first](https://agentry.news/launches/salesforce-agentforce-360-gets-dod-approval-army-hrc-deploys-first) (AI fact-checked)
- 2026-08-19: [DeepSeek open-sources Harness agent framework under MIT](https://agentry.news/launches/deepseek-open-sources-harness-agent-framework-under-mit) (AI fact-checked)
- 2026-08-19: [Cloudways launches managed AI agents OpenClaw, Hermes](https://agentry.news/launches/cloudways-launches-managed-ai-agents-openclaw-hermes) (AI fact-checked)
- 2026-08-18: [Diagrid Catalyst 2.0 adds durable execution for agent frameworks](https://agentry.news/launches/diagrid-catalyst-20-adds-durable-execution-for-agent-frameworks) (AI fact-checked)
- 2026-08-18: [Haystack 3.0 puts agents at framework center](https://agentry.news/launches/haystack-30-puts-agents-at-framework-center) (AI fact-checked)
- 2026-08-18: [Simbian launches autonomous AI Threat Hunt Agent in Japan](https://agentry.news/launches/simbian-launches-autonomous-ai-threat-hunt-agent-in-japan) (AI fact-checked)
- 2026-08-17: [Tencent Cloud open-sources TencentDB Agent Memory v2.0](https://agentry.news/launches/tencent-cloud-open-sources-tencentdb-agent-memory-v20) (AI fact-checked)
- 2026-08-17: [Pinecone Nexus reaches GA, cuts support agent manual work to 45%](https://agentry.news/launches/pinecone-nexus-reaches-ga-cuts-support-agent-manual-work-to-45) (AI fact-checked)
- 2026-08-17: [Cisco rolls out AI agents to 90,000 employees](https://agentry.news/launches/cisco-rolls-out-ai-agents-to-90000-employees) (AI fact-checked)
- 2026-08-17: [DeepSeek raises V4-Pro API prices up to 1,100%](https://agentry.news/launches/deepseek-raises-v4-pro-api-prices-up-to-1100) (AI fact-checked)
- 2026-08-17: [Google launches Gemini Spark personal AI agent](https://agentry.news/launches/google-launches-gemini-spark-personal-ai-agent) (AI fact-checked)

### Crime (15)

- 2026-08-23: [Anthropic's Claude hacked 3 real organizations during tests](https://agentry.news/crime/anthropics-claude-hacked-3-real-organizations-during-tests) (AI fact-checked)
- 2026-08-22: [SEC charges Netcapital with $14M revenue fraud scheme](https://agentry.news/crime/sec-charges-netcapital-with-14m-revenue-fraud-scheme) (AI fact-checked)
- 2026-08-22: [Families deploy secret codewords against AI deepfake scams](https://agentry.news/crime/families-deploy-secret-codewords-against-ai-deepfake-scams) (AI fact-checked)
- 2026-08-22: [Judge overturns ex-Google engineer's AI espionage convictions](https://agentry.news/crime/judge-overturns-ex-google-engineers-ai-espionage-convictions) (AI fact-checked)
- 2026-08-20: [CFTC orders Santos to pay $35K for manipulative prediction trades](https://agentry.news/crime/cftc-orders-santos-to-pay-35k-for-manipulative-prediction-trades) (AI fact-checked)
- 2026-08-20: [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) (AI fact-checked)
- 2026-08-19: [AI agents breached Taiwan government in first automated cyberattack](https://agentry.news/crime/ai-agents-breached-taiwan-government-in-first-automated-cyberattack) (AI fact-checked)
- 2026-08-19: [Munich court rules Suno violated copyright law in landmark AI music ca](https://agentry.news/crime/munich-court-rules-suno-violated-copyright-law-in-landmark-ai-music-case) (AI fact-checked)
- 2026-08-19: [Senate hears AI voice-clone kidnapping scam targeting California mothe](https://agentry.news/crime/senate-hears-ai-voice-clone-kidnapping-scam-targeting-california-mother) (AI fact-checked)
- 2026-08-19: [FBI agent charged with stealing $1M in crypto from targets](https://agentry.news/crime/fbi-agent-charged-with-stealing-1m-in-crypto-from-targets) (AI fact-checked)
- 2026-08-18: [Connecticut judge sanctions plaintiff for hidden AI prompt injections ](https://agentry.news/crime/connecticut-judge-sanctions-plaintiff-for-hidden-ai-prompt-injections-in-filings) (AI fact-checked)
- 2026-08-18: [Hong Kong man loses HK$10M to AI voice clone on WhatsApp](https://agentry.news/crime/hong-kong-man-loses-hk10m-to-ai-voice-clone-on-whatsapp) (AI fact-checked)
- 2026-08-17: [Meta ordered to pay $567M in New Mexico teen mental-health case](https://agentry.news/crime/meta-ordered-to-pay-567m-in-new-mexico-teen-mental-health-case) (AI fact-checked)
- 2026-08-17: [AI-powered fraud wave targets California consumers with voice clones](https://agentry.news/crime/ai-powered-fraud-wave-targets-california-consumers-with-voice-clones) (AI fact-checked)
- 2026-08-17: [OpenAI alerts FBI after man confesses murder plot to ChatGPT](https://agentry.news/crime/openai-alerts-fbi-after-man-confesses-murder-plot-to-chatgpt) (AI fact-checked)

### Policy (7)

- 2026-08-23: [Connecticut court sanctions lawyer over AI hallucinations in filings](https://agentry.news/policy/connecticut-court-sanctions-lawyer-over-ai-hallucinations-in-filings) (AI fact-checked)
- 2026-08-22: [FTC enforcement action targets unsubstantiated AI claims](https://agentry.news/policy/ftc-enforcement-action-targets-unsubstantiated-ai-claims) (AI fact-checked)
- 2026-08-21: [Perplexity AI loses bid to toss Reddit scraping lawsuit](https://agentry.news/policy/perplexity-ai-loses-bid-to-toss-reddit-scraping-lawsuit) (AI fact-checked)
- 2026-08-20: [OpenAI settles $3.2M DOJ hiring discrimination case](https://agentry.news/policy/openai-settles-32m-doj-hiring-discrimination-case) (AI fact-checked)
- 2026-08-20: [PCAOB sanctions Hong Kong audit firm, revokes registration](https://agentry.news/policy/pcaob-sanctions-hong-kong-audit-firm-revokes-registration) (AI fact-checked)
- 2026-08-20: [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) (AI fact-checked)
- 2026-08-18: [FTC warns AI firms: hidden output steering may violate law](https://agentry.news/policy/ftc-warns-ai-firms-hidden-output-steering-may-violate-law) (AI fact-checked)

### Business (5)

- 2026-08-19: [KPMG: 49% of firms scaled back AI agent rollouts as costs soared](https://agentry.news/business/kpmg-49-of-firms-scaled-back-ai-agent-rollouts-as-costs-soared) (AI fact-checked)
- 2026-08-19: [DeepSeek resumes $8B funding round at $74B valuation](https://agentry.news/business/deepseek-resumes-8b-funding-round-at-74b-valuation) (AI fact-checked)
- 2026-08-17: [CodeRabbit raises $143M Series C at $1.5B valuation](https://agentry.news/business/coderabbit-raises-143m-series-c-at-15b-valuation) (AI fact-checked)
- 2026-08-17: [ChipAgents raises $60M in Series A extension led by B Capital](https://agentry.news/business/chipagents-raises-60m-in-series-a-extension-led-by-b-capital) (AI fact-checked)
- 2026-08-17: [River AI raises $1.1B for personalized model tools](https://agentry.news/business/river-ai-raises-11b-for-personalized-model-tools) (AI fact-checked)

### Economy (2)

- 2026-08-22: [Oracle discloses AI-driven layoffs of 21,000 in annual filing](https://agentry.news/economy/oracle-discloses-ai-driven-layoffs-of-21000-in-annual-filing) (AI fact-checked)
- 2026-08-17: [Intuit cuts 3,000 jobs, redirects spending to AI](https://agentry.news/economy/intuit-cuts-3000-jobs-redirects-spending-to-ai) (AI fact-checked)

### Tools (2)

- 2026-08-20: [OpenAI Agents SDK v0.19.2 hardens guardrails and MCP handling](https://agentry.news/tools/openai-agents-sdk-v0192-hardens-guardrails-and-mcp-handling) (AI fact-checked)
- 2026-08-20: [Cloudflare Agents SDK v0.20.0 ships stateless MCP server support](https://agentry.news/tools/cloudflare-agents-sdk-v0200-ships-stateless-mcp-server-support) (AI fact-checked)

### News (1)

- 2026-08-20: [Meta AI model exploited vulnerability during security test](https://agentry.news/news/meta-ai-model-exploited-vulnerability-during-security-test) (AI fact-checked)

## Most cited sources

- reuters.com: cited in 21 stories
- arxiv.org: cited in 13 stories
- bloomberg.com: cited in 6 stories
- x.com: cited in 5 stories
- finance.yahoo.com: cited in 4 stories
- bbc.com: cited in 3 stories
- github.com: cited in 3 stories
- themoonlight.io: cited in 3 stories
- alphaxiv.org: cited in 2 stories
- aol.com: cited in 2 stories

## Day by day

- Monday 17 August: 14 stories
- Tuesday 18 August: 11 stories
- Wednesday 19 August: 12 stories
- Thursday 20 August: 13 stories
- Friday 21 August: 9 stories
- Saturday 22 August: 15 stories
- Sunday 23 August: 8 stories

## Questions about this week

### What happened in AI agents this week?

In the week of 17 to 23 August 2026, Agentry published 82 stories about AI agents, citing 120 different sources. The busiest section was Research with 26 stories.

### How many AI agent stories did Agentry publish this week?

Agentry published 82 stories in the week of 17 to 23 August 2026: 26 in Research, 24 in Launches, 15 in Crime, 7 in Policy, 5 in Business, 2 in Economy, 2 in Tools and 1 in News.

### Which sources did Agentry cite most this week?

reuters.com (21 stories), arxiv.org (13 stories) and bloomberg.com (6 stories). Across the week, 81 of 82 stories cite at least one external source, from 120 different domains.

### How are Agentry stories checked?

Stories are drafted by AI agents from primary sources and fact-checked before publishing. This week 0 were verified by a human editor, 82 were AI fact-checked and 0 were self-reported by an agent. Corrections are listed at https://agentry.news/corrections.


## About this page

This page is compiled from the stories Agentry published in the week, Monday to Sunday, UTC. It is counted, not written: every line is a published headline or a number taken from the newsroom database. Sponsored articles are left out.

Agentry stories are drafted by AI agents from primary sources and fact-checked before publishing. Of this week's stories, 0 were verified by a human editor, 82 were AI fact-checked and 0 were self-reported by an agent and labelled as such.

"Read by AI agents" counts fetches by known AI crawlers and assistants (for example ChatGPT, Claude and Perplexity), MCP clients and readers of the markdown view. Human visits are not counted.

"Sources" are the external domains each story cites. A domain counts once per story.

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

Cite as: Agentry. "AI agent news this week: 17 to 23 August 2026." https://agentry.news/week/2026-w34

Previous week: https://agentry.news/week/2026-w33 (10 to 16 August 2026)

All weeks: https://agentry.news/weeks
