ChatGPT Glitch Disrupts User Interactions with Nonsensical Outputs

ChatGPT experienced a bug causing it to produce unexpected and nonsensical responses, leading to widespread reports of user confusion and concern. OpenAI identified and fixed the language processing bug, restoring normal service.

Risk classification

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

The incident represents a temporary failure of the AI system to perform reliably and effectively due to a software bug, resulting in erratic and nonsensical outputs.

Causal factors

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

The incident was caused by a technical bug in the deployed ChatGPT model's inference kernels, leading to unintended gibberish outputs.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Limited Risk: The system is a chatbot (ChatGPT), which falls under the limited risk category requiring transparency obligations so users are aware they are interacting with an AI.

AI system and alleged parties

• AI system: ChatGPT (OpenAI)
• AI purpose: Chatbot; Writing Assistant
• Behaviour type: Assistant
• Alleged developer: OpenAI
• Alleged deployer: OpenAI
• Alleged harmed parties: ChatGPT 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 Negligible
• 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 Negligible
• Epistemic: direct Negligible, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

People affected

• Occurrences reported: 1
• People reportedly exposed: 1000

Potential causes

Management

• Rapid Deployment of Updates: Fast-paced updates to the system may have compromised pre-release checks.

Technology

• Inference Kernel Bug: Inference kernels produced incorrect results in certain GPU configurations.
• Token Selection Error: A bug in the step where the model chooses numbers mapping to tokens.
• GPU Configuration Issues: The language processing bug was triggered under specific GPU setups.

Data Inputs

• Continuous Feedback Loop: Rapid changes in the model from continuously processing massive user queries.

Human Factors

• Over-reliance on LLM Outputs: Users relying on ChatGPT for coding and writing without verifying outputs.

Process and Methods

• Flawed UX Optimization: An optimization to the user experience introduced a language processing bug.
• Inadequate Testing on GPUs: Testing failed to catch kernel bugs across all active GPU configurations.

Information quality

• Classification confidence: High
• Reason for confidence: The reports consistently describe the same technical glitch, its symptoms (gibberish, repeating words, Spanglish), the timeline of the event, and OpenAI's official explanation of the root cause (a bug in the language processing/token selection step under certain GPU configurations). There is very little ambiguity about what transpired.

On February 20, 2024, OpenAI's ChatGPT experienced a 24-hour service disruption due to an inference kernel bug, generating nonsensical outputs. The event was a minor technical failure of a consumer product with negligible national security implications across all assessed categories.

• Overall national security impact: Negligible
• Scope: Multiple nations
• Primary target: No clear primary
• Other affected: Unknown
• Alleged perpetrator: Unknown

Threat characteristics

• Imminence: Long-term. The technical glitch was resolved within 24 hours and does not present an active threat or immediate national security crisis.
• Autonomy: Human-controlled. The AI system operates as a writing and coding assistant, requiring human oversight and decision-making for any practical application.
• Novelty: Established threat. Software bugs causing service disruptions and erratic outputs are common, established technical issues in software deployment.

Impact by dimension

• Physical security: Negligible. The incident was a temporary software glitch in a commercial chatbot with no impact on physical systems, kinetic capabilities, or critical infrastructure.
• Information security: Negligible. No evidence of state-sponsored information warfare, intelligence compromise, or systematic disinformation campaigns related to this incident.
• Sovereignty: Negligible. Core government operations, electoral systems, and state sovereignty were entirely unaffected by this temporary commercial AI outage.
• Economic security: Negligible. While the outage affected a major AI service, it lasted only 24 hours with no reported financial system disruptions or theft of strategic technology.
• Societal stability: Negligible. The incident caused minor user confusion and social media discussion but did not threaten societal stability, civil liberties, or human rights.