AI and 911 Call Systems: A New Ally or a Hidden Risk?

In today’s world, where the unexpected can strike at any moment, the systems designed to protect people – 911 call centers – are facing unprecedented challenges. From natural disasters to sophisticated cyberthreats, the risks to public safety are evolving fast. For example, a devastating wildfire in Lahaina, Maui, in August 2023 destroyed much of the town, causing significant loss of life and property. The call volume overwhelmed 911 dispatchers. A ransomware attack on Change Healthcare in February 2024 caused massive disruptions in the U.S., including outages with some 911 call centers. Yet, these critical lifelines are often hampered by outdated technology, limited resources, and the delicate balance between ensuring security and preserving personal freedoms. Behind each call for help is a human life at stake. As these threats grow more complex, so too does the responsibility of those tasked with responding. The current environment is a delicate and dynamic one, where the consequences of failure can be devastating, whether from human error or the limits of precautionary measures. 

The Principles of Communication are vital during routine and emergency situations involving the reporting and assessment of an event to the dispatch and the utilization and response of law enforcement and public safety professionals. The most effective use of resources for appropriate response and risk mitigation can be hindered by poor communication among and between the public and first responders. Unfortunately, the challenges of poor interoperability and means of communication among law enforcement and public safety professionals continue to be a national problem, with over two decades of intermittent progress, incomplete measures, and sporadic attention. Additionally, the complexity and uncertainty that are characteristic of crisis situations add to the “fog of war” and may also delay, impede, or otherwise negatively impact the communication inherent in successful public safety efforts. 

And now Artificial Intelligence (AI)… 

One facet of homeland security and public safety subject to increased efficiencies and risks associated with advancements in technology is the nation’s 911 call systems. These systems play a pivotal role in ensuring swift and effective emergency responses. The integration of AI technologies can significantly enhance these systems. AI and its impact on critical domains are crucial to homeland security. The public relies on the 911 call system during emergencies and crises. 

It is not a far stretch to imagine the deafening sounds of a fire truck racing to a three-alarm fire, the arrival of a police squad car at the scene of an assault, or the swift departure of an ambulance carrying a heart attack victim to the nearest emergency room. These scenarios are initiated and depend on a reliable and trustworthy 911 call system. These three numbers (911) and subsequent emergency response can be the difference between life and death. 

However, this lifeline is plagued with problems due to antiquated systems, funding shortfalls, and inadequate staffing, training, and capabilities. Shortcomings of the 911 call system cause service delays and outages. These failures occur across the country and add to the challenges of public safety and preparedness. 

The Pros and Cons of AI for First Responders and 911 Systems 

The integration of AI technologies can significantly enhance these systems and benefit society. Some pros and cons of using AI for first responders and 911 emergency call systems, along with examples of the challenges and benefits, are listed below.

Pros: 

1. Faster Response Times – AI can process information rapidly, helping dispatchers prioritize urgent calls and allocate resources efficiently. Example: Real-time location tracking and video capabilities assist responders in reaching emergency scenes faster. 
2. Automated Call Routing, Triage, and Decision Support – AI algorithms can intelligently route emergency calls to the nearest dispatch center based on location data. This reduces response time and ensures that help reaches the scene promptly. AI algorithms can assess the severity of calls and provide recommendations to dispatchers. Example: AI-based tools help identify life-threatening situations and guide responders accordingly. 
3. Natural Language Processing (NLP) and Language Translation – AI-powered NLP models can transcribe and analyze callers’ speech, extracting critical information even when the caller is distressed or unable to communicate clearly. AI can instantly translate calls in different languages, bridging communication gaps. Example: A dispatcher can communicate effectively with non-English-speaking callers. 
4. Predictive Analytics – AI analyzes historical data to predict emergency trends, allocate resources proactively, and optimize emergency services’ deployment. Example: Predictive models anticipate spikes in call volume during natural disasters. 

Innovation in AI technologies and the increased efficiencies that follow are not risk-free. There are potential barriers, concerns, and considerations. 

Cons: 

1. Bias and Fairness – AI systems may inherit biases from training data, affecting decision-making. AI systems must be trained on datasets to avoid biases related to race, gender, or socioeconomic status. Biased algorithms could inadvertently impact emergency response decisions. Example: An AI algorithm might inadvertently prioritize certain neighborhoods over others. 

1. Privacy Concerns – Balancing the need for efficient emergency services with privacy rights is crucial. AI processes sensitive information during calls, raising privacy issues. The systems should handle sensitive information securely to protect the public and promote trust. Example: Balancing data collection for emergency response with privacy rights. 

1. Overreliance and Errors – Relying solely on AI can lead to mistakes or missed critical details. Example: An AI system may misinterpret a caller’s distress level due to speech nuances. 

1. Human Interaction and Empathy – AI lacks human empathy and emotional understanding. Example: Callers may need emotional support during distressing situations. 

1. Trust and Transparency – Community members may be skeptical of AI-driven responses. Example: Use of AI could lead to further mistrust in law enforcement and public safety agencies 

These are samples of the benefits and challenges for the community posed using AI-driven 911 systems. Striking the right balance between AI assistance, data-driven decision-making and human judgment is crucial. Overreliance on AI could impact critical decision-making. AI offers significant advantages for managing call volumes and emergency response and improving efficiency. However, careful implementation and ongoing evaluation are essential to maximize benefits while addressing potential drawbacks. 

Although few states (e.g., Colorado, Maryland, Missouri, Oregon, South Carolina, Texas, and Virginia) have established clear AI regulatory frameworks for emergency call centers, some progress has been made in addressing the vulnerabilities of an AI-infused call system. For example, some actions and regulations undertaken by public safety agencies and police professionals aim to protect communities from the malicious use of AI in emergency dispatch services. 

The identification and disruption of malicious AI actors is vital. Organizations like OpenAI collaborate with partners to detect and disrupt state-affiliated threat actors using AI for cyberattacks. Mitigation efforts include terminating accounts associated with malicious actors, which helps prevent misuse of AI services. Prescribed guidelines for AI bots in 911 centers are essential for balancing efficiency and accuracy in AI use while addressing bias concerns and potential errors in AI responses. 

Emergency Services as a Likely Target for Cyberattacks, Warns DHS 

Despite no widespread reports of 911 call systems being directly hijacked by malicious AI, concerns about the potential risks exist. Akin to the nation’s critical infrastructure, the nation’s 911 emergency services and call systems are also potential targets and vulnerable to exploitation and cyberattacks. Following is a hypothetical scenario and possible mitigation steps to explore. 

Scenario: AI-Driven 911 System Hijacking 

• The Setup – A mid-sized city with an AI-enhanced 911 system relies on machine learning algorithms to process emergency calls. The AI system is trained to recognize distress signals, prioritize calls, and dispatch appropriate responders. 

• The Attack – Malicious actors exploit vulnerabilities in the AI system: 

• Adversarial Inputs – They craft fake emergency calls designed to confuse the AI. 

• Data Poisoning – They manipulate training data to bias the AI’s decision-making. 

• Model Tampering – They compromise the AI model itself. 

• Consequences: 

• False Prioritization – The AI misclassifies critical calls, leading to delayed responses for life-threatening emergencies. 

• Resource Misallocation – Responders are dispatched to non-emergencies, leaving genuine emergencies unattended. 

• Chaos and Panic – The public loses trust in the 911 system, causing panic during real crises. 

• Attack Examples: 

• Swatting Attacks – Malicious actors use AI-generated calls to falsely report emergencies (e.g., bomb threats, active shooters) at specific locations. Responders rush to the scene, wasting resources. 

• Data Poisoning – By subtly altering training data, attackers bias the AI to ignore certain types of calls (e.g., domestic violence) or prioritize others (e.g., high-profile areas). 

• Mitigation: 

• Robust Testing – Regularly test the AI system against adversarial inputs. 

• Human Oversight – Maintain human dispatchers alongside AI to catch anomalies. 

• Secure Training Data – Prevent data poisoning by carefully curating training datasets. 

This hypothetical scenario is not much different than the potential reality faced by the law enforcement and first responder profession and the public. The challenges for improving operability have been outlined and, like others facing homeland security professionals, the need for additional funding is common. The annual spending on 911 call systems varies by state and locality. 

National View: 911 Data 

In 2015, 40 states and the District of Columbia collectively spent approximately $3.4 billion on 911 services. The National 911 Annual Report and the National 911 Profile Database provide more recent and specific data, including an interactive version of the most recent data. 

The National 911 Annual Report is a collaborative effort between the National 911 Program and the National Association of State 911 Administrators (NASNA). Each year, they collect comprehensive data related to 911 services, including funding, revenue, text-to-911 implementation, and progress toward Next Generation 911 (NG911). This data is voluntarily submitted by states and compiled in the National 911 Profile Database. 

The report analyzes trends and provides segmented information for state 911 leaders, legislators, and policymakers to make informed decisions about emergency services. The latest report covers 2021 data, and previous reports are available for reference. The Profile Database captures details about a state’s 911 operations, protocols, and progress, helping enhance emergency response nationwide. 

The 2021 National 911 Annual Report revealed several noteworthy findings across statewide plans. Thirty-three states reported having a statewide Next Generation 911 (NG911) plan. NG911 aims to enhance emergency communication infrastructure and services along with adoption of the Emergency Services IP Network (NG911 ESInet). Over 2,000 Public Safety Answering Points (PSAPs) across 46 states reported using an Emergency Services IP Network (ESInet). ESInets facilitate efficient call routing and data exchange in emergencies. Also, in 38 states, nearly 600,000 texts-to-911 were received during the calendar year. The Text-to-911 feature provides an alternative communication channel for those unable to make voice calls.  

In summary, artificial intelligence is a double-edged sword in the realm of public safety and security. On one side, it empowers emergency responders, fortifies defenses, and enhances the ability to predict and respond to crises. On the other, it carries risks that, if not carefully managed, could compromise the systems it seeks to strengthen. To fully harness the power of AI while protecting society, community stakeholders must ensure its responsible implementation, prioritize transparency, and uphold rigorous ethical standards. The balance between innovation and caution will determine whether AI serves as a community’s greatest ally or a hidden danger in the shared mission to protect human lives.