Google Android Earthquake Alerts System Reportedly Underestimated Kahramanmaraş Earthquake and Failed to Send Urgent Warnings to Millions

During the February 6, 2023 Kahramanmaraş earthquake sequence, Google's Android Earthquake Alerts system reportedly estimated the initial magnitude-7.8 earthquake at only magnitude 4.5–4.9. Just 469 users reportedly received its highest-level "Take Action" warning, while approximately 500,000 received lower-level notices; researchers later estimated that roughly 10 million people could have received urgent warnings providing up to 35 seconds to seek safety.

Risk classification

• Primary risk domain: 7 AI system safety, failures, & limitations
• Primary risk subdomain: 7.3 Lack of capability or robustness

The AI system failed to perform robustly under real-world conditions because its magnitude detection algorithm miscalculated a major 7.8 earthquake as a minor 4.5-4.9 event.

Additional risk subdomains

• 5.1 Overreliance and unsafe use: Experts expressed concern that overreliance on Google's supplemental system might discourage countries from building their own national warning systems.

Causal factors

• Entity: AI
• Intent: Unintentional
• Timing: Post-deployment

The failure was caused by an algorithmic error in Google's deployed earthquake alert system, which unintentionally miscalculated the earthquake's magnitude.

EU AI Act risk tier

• Risk tier: 2 High Risk

High Risk: The system is used in critical infrastructure or public safety contexts (earthquake early warning) where failure directly impacts life and limb, as noted by experts who stated Google has a responsibility to follow through on services related to life and limb.

AI system and alleged parties

• AI system: Android Earthquake Alerts System (Google)
• AI purpose: Threat Detection
• Behaviour type: Autonomous
• Alleged developer: Google, Earthquake early warning system developers
• Alleged deployer: Google, Earthquake early warning system deployers
• Alleged harmed parties: Public safety, People exposed to earthquake hazards, General public of Turkey, General public, Android users in earthquake-affected areas of Turkey, Android users in earthquake-affected areas, Android users

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Negligible. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Severe
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Negligible, indirect Negligible
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Negligible, indirect Negligible
• Differential treatment: direct Negligible, indirect Negligible
• Civil rights: direct Negligible, indirect Negligible
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Negligible, indirect Negligible
• Psychological: direct Negligible, indirect Substantial
• Epistemic: direct Negligible, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Physical

Reported: The report explicitly describes massive physical harm, including over 55,000 fatalities, from the earthquake, but notes the AI system's failure to warn residents may have contributed to the loss of life.

Indirectly caused: The failure of the system to send high-level alerts to an estimated 10 million people deprived them of up to 35 seconds of warning, which could have allowed them to escape collapsing buildings, indirectly contributing to the 55,000 deaths and over 100,000 injuries.

Psychological

Reported: The report describes survivors experiencing severe grief and trauma from losing family members, though it does not explicitly quantify psychological harm caused by the AI failure itself.

Indirectly caused: Survivors like Funda, who lost 25 family members, experienced profound trauma and grief, which was compounded by the failure of the warning system they expected to protect them.

Inferred additional harm: Widespread psychological trauma, anxiety, and PTSD are highly likely among the millions of residents in the affected zone who did not receive the expected life-saving alerts.

People affected

• Occurrences reported: 1
• People reportedly harmed: 55000
• People reportedly exposed: 10000000

Potential causes

Management

• Defensive Public Communication: Management insisted the system worked without providing verifiable data.
• Overreliance on Supplemental Tech: Positioned experimental tech as a core safety service without validation.

Technology

• Flawed Magnitude Estimation: Algorithm underestimated the 7.8 magnitude quake as a 4.5 to 4.9 event.
• Inadequate Alert Tiering: Low-level alert did not bypass Do Not Disturb mode for sleeping users.

Data Inputs

• Crowdsourced Sensor Data: Accelerometers struggled to provide data for accurate large-scale estimation.

Human Factors

• Active Do Not Disturb Mode: Sleeping users did not hear the low-level alert due to active DND settings.

Process and Methods

• Poor Large Quake Calibration: Algorithms lacked calibration to accurately estimate massive scale earthquakes.
• Insufficient Public Reporting: Google failed to share performance data transparently for two years.

Regulatory Environment

• Lack of Official Standards: No regulatory body enforced performance or transparency standards on Google.

Information quality

• Classification confidence: High
• Reason for confidence: The reports provide detailed technical explanations from both Google's own scientific publications and independent seismic experts, clarifying exactly why the algorithm failed (miscalculating a 7.8 magnitude as 4.5-4.9) and the precise number of alerts sent (469 instead of 10 million).
• Ambiguities identified: The exact number of lives that could have been saved had the system functioned perfectly is not quantified, as it depends on human reaction times and building structural integrity.
• Alternative interpretations: The failure could be viewed purely as a communication network limitation during a disaster, but Google's own scientific paper confirmed the primary failure was algorithmic magnitude miscalculation.

Google's Android Earthquake Alerts system failed to warn millions of users during the 2023 Turkey earthquake due to an algorithmic miscalculation of the disaster's magnitude. While resulting in severe public safety consequences, the incident represents a minor direct threat to national security, highlighting the limitations of relying on autonomous consumer-grade systems for critical emergency alerts.

• Overall national security impact: Minor
• Scope: Single nation
• Primary target: Turkey
• Alleged perpetrator: Unknown

Threat characteristics

• Imminence: Long-term. The failure represents an ongoing technological capability and reliability concern for public safety systems rather than an active crisis.
• Autonomy: Full autonomy. The AI system autonomously processes real-time sensor data and decides whether to trigger high-level emergency alerts without human intervention.
• Novelty: Evolved capability. The incident demonstrates a failure in an evolved, crowdsourced smartphone-based seismic detection capability rather than a completely unprecedented technology.

Impact by dimension

• Physical security: Minor. The algorithmic failure of a supplemental consumer earthquake alert system during a major disaster led to a lack of timely warnings for millions, representing a public safety issue but minor direct national security impact.
• Information security: Negligible. No information warfare, intelligence compromise, or systematic information manipulation occurred in this incident.
• Sovereignty: Negligible. No direct threat to state authority, territorial control, or core government operations was observed, though it highlights reliance on foreign tech for public safety.
• Economic security: Negligible. The incident represents an algorithmic limitation of a consumer product rather than a strategic threat to economic stability or critical technological assets.
• Societal stability: Negligible. The event did not involve mass surveillance, algorithmic oppression, or systematic violations of civil liberties.