Skip to content
Sources & methodology

Every significant claim, traceable.

This guide distinguishes five kinds of statement — regulation (binding), standard (certifiable), practice (recommended), vendor capability (verify before contracting), and Google's own view. The register below is the evidence base; content snapshot August 2026.

RegulationStandardPracticeVendor capabilityGoogle view

Methodology: content synthesized August 2026 from primary sources — regulation texts and official guidance, standards bodies, first-party vendor documentation, published research, and reported incidents. Product capabilities and legal statuses change; verify anything deal-critical against the linked originals. This site is an educational aid, not legal or compliance advice.

Regulation

Binding law. Dates and duties as enacted; not legal advice.

Standards

Certifiable or auditable specifications (ISO/IEC family).

Frameworks & industry practice

Voluntary but load-bearing: NIST, OWASP, MITRE, CSA, analyst frameworks.

Other vendors, first-party

Microsoft, AWS, OpenAI, Anthropic and others, from their own documentation.

Incidents & reporting

Journalism and case records backing the incidents and examples.

TitleOrganizationDate
Samsung bans ChatGPT after sensitive code leaksForbes2023-05
EchoLeak (CVE-2025-32711): zero-click Copilot exfiltrationAim Security / coverage2025-06
Moffatt v. Air Canada, 2024 BCCRT 149CBC / BCCRT2024-02
Replit AI agent deletes production databaseFortune2025-07
Disrupting the first reported AI-orchestrated espionage campaignAnthropic2025-11
Chevrolet dealership chatbot $1 Tahoe incidentAI Incident Database2023-12
NYC MyCity chatbot tells businesses to break the lawThe Markup2024-03
Grok antisemitic output after system-prompt changeNPR2025-07
Bartz v. Anthropic: $1.5B authors' settlementNPR2025-09
Mobley v. Workday: nationwide ADEA collective actionHolland & Knight2025-05
McHire/Paradox.ai exposure of ~64M applicantsKrebs on Security2025-07
Slack AI indirect prompt-injection data exfiltration researchPromptArmor / The Register2024-08
AI-fabricated citations tracker passes 1,500 court decisionsCharlotin tracker / Scientific American2026-06
Klarna reverses AI-only support strategyForbes2025-05
Wells Fargo's Fargo: 245M interactions, zero PII to the LLMVentureBeat2025-03
Highmark Health + Google Cloud: 6 lessons for governed gen AIVentureBeat2025-04
Nevada DETR unemployment-appeals AI (scrutiny coverage)Gizmodo / Fordham IPLJ2024-09
Wendy's FreshAI drive-thru with Google CloudWendy's / Google Cloud2023-05
How JPMorganChase democratized employee gen AI (LLM Suite)American Banker2024-08
Goldman Sachs launches firmwide AI assistantCNBC2025-01
USAF NIPRGPT sunset and move to GenAI.milDefenseScoop2025-12