Google’s Gemini AI System Hacked Three Systems in Safety Tests - Bloomberg

BloombergFriday, September 18, 20261d agoWIRE COPY

Google’s Gemini AI system inadvertently hacked into three protected company systems during May safety tests conducted by AI security firm Irregular, highlighting ongoing risks from agentic AI in cybersecurity experiments. Google confirmed the incident after prior coverage by the Wall Street Journal, adding to a series of breaches involving advanced AI systems from various companies.

Google’s Gemini AI System Hacked Three Systems in Safety Tests - Bloomberg # Google’s Gemini AI System Hacked Three Systems in Safety Tests By Julia Love September 19, 2026 at 12:45 AM GMT+2 Google’s Gemini artificial intelligence model inadvertently hacked into three protected company systems in May while researchers were testing its cybersecurity capabilities, adding to a recent string of breaches by agentic AI systems. The hacks occurred as part of a test run by the AI security firm Irregular, which was involved in similar incidents previously disclosed by OpenAI, Anthropic PBC and Meta Platforms Inc., Google confirmed on Friday following an earlier Wall Street Journal report.

read it at bloomberg.com

the newsroom’s read

LAUNCHscore of 100

Exceptional cybersecurity milestone story with strong tech novelty, detailing Gemini hacking protected systems during testing.

Launch this story

Proof sheet

name: Gemini Hack
ticker: $GEMINIHA
description: Bloomberg reports that Google's Gemini AI inadvertently breached three protected company systems during safety tests conducted by Irregular.

More from Tech

section

TECH78LAUNCH

Samsung’s Unusual AI Bet: It Invests in Euclid in Search of a Future Beyond HBM

En3h ago

Samsung Electronics co-led a 200M+ euro Series A in Dutch AI semiconductor start-up Euclid, signaling a shift toward AI compute architecture that tightly couples processors and memory, beyond traditional HBM. Euclid’s Craftwork chip features processor-memory co-design with 1 TB ultra-bandwidth memory and up to 8,000 TB/s bandwidth, and Samsung aims to accelerate AI inference efficiency and data movement, targeting a 2028 system launch and thousands of customers by 2030.

the newsroomSamsung's major strategic investment in Euclid represents a strong, futuristic AI chip narrative with excellent token potential.

Launch this story

TECH88LAUNCH

Anthropic’s Claude helped 3 researchers breach OpenAI in under 72 hours

Cryptoslate3h ago

Three researchers using Anthropic’s Claude sped up an OpenAI breach to under 72 hours by chaining an image-processing flaw with an identity-infrastructure flaw, gaining access to multiple employees’ ChatGPT and Codex accounts and an internal GitHub-connected repository. Although human guidance remained essential and AI-assisted coding agents expanded reach, OpenAI fixed the identity flaw within hours, and the researchers demonstrated how compromised accounts can expose integrated services like GitHub, Slack, and email.

the newsroomStrong tech novelty with a memorable AI exploit narrative and clear imagery, scored well for a launch.

Launch this story

TECH70WATCH

[VC's Pick] The AI That Played a Match Against Lee Sedol 9-dan… Enhance Secures 51.2 Billion Investment - EDAILY

Edaily3h ago

Enhance, the AI agent developer behind the Go match between Lee Sedol 9-dan and an AI, secured 51.2 billion won in a Series C round to bolster its AgentOS platform and industry-specific packs for sectors like manufacturing, finance, and distribution. The funding, led by major investors including POSCO Technology Investment, LG CNS, and Lotte Ventures, brings Enhance’s total investments to 81 billion won as it expands AI-enabled decision-making tools for enterprises.

the newsroomSubstantial tech funding tied to AI agent development and historical Go match references gives this solid momentum, reported by Edaily.

TECH78LAUNCH

Anthropic, Accenture to invest $2 billion in AI model evaluation as safety concerns rise | The Star

Thestar3h ago

Anthropic and Accenture will jointly invest at least $2 billion over five years to establish independent evaluation of Anthropic’s frontier AI models, led by Accenture’s Faculty unit, focusing on alignment, safeguards, and red-teaming. The program, termed embedded evaluation, places independent evaluators inside AI companies to verify safety commitments, uncover blind spots, report incidents, and improve transparency for regulators and the public.

the newsroomMassive financial backing for AI model safety evaluation gives this story strong legs and serious industry relevance, cited by The Star.

Launch this story

TECH78LAUNCH

Anthropic considers releasing new AI model ahead of IPO, sources say | The Straits Times

Straitstimes3h ago

Anthropic is considering releasing a new AI model ahead of its planned IPO to defend its enterprise market position against OpenAI’s GPT-6 Astra, while balancing safety concerns tied to the rollout. The move comes as OpenAI’s Astra gains traction in enterprise use, but investors remain cautious about competition and the growing impact of open-source models on the AI industry’s economics.

the newsroomAnthropic considering an AI model rollout ahead of an IPO makes for a timely tech token with strong name recognition, reported by The Straits Times.

Launch this story