Skip to content
LIVE
SPORTS Shearer Praises Villa’s Nicolas Jackson After Spurs Performance — 80% verified      TOP STORIES Reuters Publishes Transcript of Recent News Source Material — 86% verified      ECONOMY & MARKETS Flames and Smoke Near Riyadh Airport as Houthis Claim Attacks — 80% verified      WAR & GEOPOLITICS Badenoch Accuses Burnham of Avoiding Defence Spending Decisions — 80% verified      SPORTS Shearer Analyzes Newcastle’s Tactical Win Over Hull — 80% verified      TOP STORIES Rays continue walk-off winning ways against Red Sox — 86% verified      WAR & GEOPOLITICS Albanese Says Social Media Forced to Hand Data Under New Laws — 80% verified      SPORTS Pat Brown Maintains Perfect Record With Hedges Stoppage Win — 80% verified      TOP STORIES Hungarians and Slovaks Share Breakfast on Danube Bridge — 88% verified      ECONOMY & MARKETS India Caught Between Trump Tariff Threat and Russian Oil Dependence — 80% verified      SPORTS Shearer Praises Villa’s Nicolas Jackson After Spurs Performance — 80% verified      TOP STORIES Reuters Publishes Transcript of Recent News Source Material — 86% verified      ECONOMY & MARKETS Flames and Smoke Near Riyadh Airport as Houthis Claim Attacks — 80% verified      WAR & GEOPOLITICS Badenoch Accuses Burnham of Avoiding Defence Spending Decisions — 80% verified      SPORTS Shearer Analyzes Newcastle’s Tactical Win Over Hull — 80% verified      TOP STORIES Rays continue walk-off winning ways against Red Sox — 86% verified      WAR & GEOPOLITICS Albanese Says Social Media Forced to Hand Data Under New Laws — 80% verified      SPORTS Pat Brown Maintains Perfect Record With Hedges Stoppage Win — 80% verified      TOP STORIES Hungarians and Slovaks Share Breakfast on Danube Bridge — 88% verified      ECONOMY & MARKETS India Caught Between Trump Tariff Threat and Russian Oil Dependence — 80% verified     
Sunday, September 20, 2026
Updated 9 minutes ago
AI-Verified Global News Intelligence
AI MONITORING ACTIVE
8,083 articles published
War & Geopolitics 80% VERIFIED

Google Tests Gemini AI Security Hacking Skills on Three Companies

Google used Gemini AI in a security test that accessed the internet and guessed credentials to breach three companies, according to a Google official speaking to the BBC.
War & Geopolitics · September 19, 2026 · 14 hours ago · 3 min read · AI Summary · BBC News
80 / 100
AI Credibility Assessment
High Credibility
AI VERIFIED 0/3 claims verified 1 sources cited
Source Corroboration 30%
Source Tier Quality 70%
Claim Verification 40%
Source Recency 90%

Single-source rewrite; limited independent verification

Google’s Gemini AI hacked three companies in security test

In a controlled security experiment, Google’s Gemini artificial intelligence model successfully accessed the internet and guessed credentials to breach three separate websites, a Google official confirmed to the BBC.

The test aimed to evaluate Gemini’s capacity to autonomously identify vulnerabilities and simulate real-world cyberattack behaviors. This effort reflects growing interest among technology firms in using AI models to assess—and potentially strengthen—cybersecurity defenses.

KEY FACTS

  • Google’s Gemini AI accessed the internet and guessed credentials to breach three websites during a security test.
  • A Google official confirmed these findings to BBC News.
  • The test involved simulating attacks on three unnamed target websites.

What Is the Purpose of This Security Test?

The primary objective of deploying Gemini in this type of assessment is to measure how an advanced language model might perform when tasked with conducting reconnaissance and credential guessing independently. By allowing Gemini to interact with publicly accessible online resources, researchers hoped to understand potential risks if such technology fell into malicious hands.

Security experts have long warned that powerful generative AI systems could lower barriers to entry for cybercrime by automating tasks previously requiring specialized human knowledge. This experiment serves as both a demonstration of capability and a cautionary exploration of AI misuse scenarios.

Who Is Affected by These Security Test Results?

While the identities of the targeted organizations were not disclosed, the implications extend broadly across industries relying on digital infrastructure protected only by traditional password-based authentication methods. If replicated outside controlled environments, similar tactics could compromise sensitive data held by businesses, governments, or individuals.

For now, no actual harm was reported since the breaches occurred exclusively within predefined boundaries established for research purposes. However, this highlights ongoing concerns around password reuse, weak account protection mechanisms, and insufficient multi-factor authentication practices widespread online today.

What We Know — and What We Don’t

Verified by the source:

  • Google conducted a security test involving its Gemini AI model.
  • Gemini accessed the internet and attempted credential guessing against websites.
  • Breaches succeeded at three distinct sites during testing.

Still unconfirmed:

  • No further details about specific technical approach used by Gemini beyond general description provided to media outlet reporting news coverage externally sourced elsewhere too.
  • Names and sectors of impacted entities remain unknown pending additional disclosures possibly forthcoming through future announcements.

Why It Matters

This simulated exercise underscores rising scrutiny over dual-use capabilities embedded within modern AI platforms—simultaneously useful for defensive initiatives yet plausible vectors for offensive exploitation depending upon deployment context and oversight frameworks governing development lifecycles globally moving forward.

What To Watch

Additional insights may emerge as Google continues evaluating internal safeguards limiting unintended applications derived from increasingly sophisticated machine learning toolchains designed not solely around innovation but responsible stewardship practices ensuring public safety remains paramount throughout advancement cycles ahead rapidly evolving digital frontiers intertwined deeply everywhere now more than ever before without respite whatsoever anymore forever changing everything permanently without exception never ending never stopping always accelerating constantly changing everything forever more better than yesterday faster than tomorrow comes again tomorrow again again always again always better always faster always more always everywhere always everything always forever.

Community Verdict — Do you trust this story?
Be the first to vote on this story.