This Month's Latest Tech News in San Antonio, TX - Wednesday April 30th 2025 Edition

San Antonio is rapidly evolving into a major player in AI innovation and ethical technology, with the city's focus on smart city initiatives and advanced education.

The upcoming Smart Cities Connect Spring Conference & Expo highlights the region's commitment to integrating AI for public safety, efficient urban management, and citizen-centered services.

At the heart of this transformation, the University of Texas at San Antonio (UTSA) has launched the College of AI, Cyber and Computing, aiming to create a talent pipeline in AI, cybersecurity, and data science - a move described as “a monumental step towards bolstering our city's position as a hub for technological innovation and security” by USAA's CIO, Amala Duggirala (UTSA AI, Cyber and Computing).

Meanwhile, San Antonio's smart city program is distinguished by ethical leadership and rapid prototyping, as Emily Royall, head of the city's emerging technology division, notes:

“Smart cities ask what problems technology solves and for whom - never pursuing technology for its own sake. This ensures taxpayer dollars are invested wisely.”

Discover more about how San Antonio sets the pace for responsible, community-informed tech adoption in Government Technology's report on the next evolution of smart cities.

A groundbreaking study by researchers at the University of Texas at San Antonio (UTSA) has spotlighted a significant cybersecurity threat in AI-assisted software development: package hallucination.

This occurs when large language models (LLMs) generate or recommend convincing - but nonexistent - software packages, inadvertently enabling “slopsquatting” attacks where malicious actors publish harmful code under these hallucinated names.

Analysis of over 2.2 million code samples revealed that open-source AI models hallucinated package names at a rate of 21.7%, compared to 5.2% for commercial models, with JavaScript code more vulnerable than Python.

The study's findings are summarized below:

Model Type	Hallucination Rate
Open Source	21.7%
Commercial (GPT-series)	5.2%

These persistent hallucinations create an ideal environment for supply chain attacks, as developers may blindly trust and install recommended dependencies.

As highlighted in Wired's in-depth report on AI code hallucinations, the risk escalates as AI-generated code becomes ubiquitous.

The UTSA team warns,

“You're placing a lot of implicit trust on the package publisher... but every time you download a package; you're potentially downloading malicious code.”

“Package hallucinations were found to be a systemic issue across all models and languages tested, including state-of-the-art commercial models.” - Joe Spracklen, UTSA

Industry experts urge verifying AI-suggested packages and improving LLM training, as detailed in Dark Reading's review of the slopsquatting threat.

For a comprehensive breakdown of UTSA's findings and recommendations, see the original UTSA study on AI threats in software development.

WM has officially unveiled its $72 million Mesquite Creek Recycling Facility in New Braunfels, Texas, bringing state-of-the-art artificial intelligence and optical sorting technology to one of America's fastest-growing regions.

As part of WM's broader $3 billion sustainability investment plan through 2026, this 110,000-square-foot facility can process up to 144,000 tons of recyclables annually, dramatically expanding recycling capacity and access for a previously underserved community.

Core innovations include 16 intelligent optical sorters and advanced sensors that efficiently recover plastics, fibers, metals, and glass, while automated conveyors boost both material quality and processing speed.

The project is central to WM's enterprise-wide goal of adding 2.8 million tons in annual recycling capacity and generating 25 million MMBtu of renewable natural gas - enough to power up to 1.7 million homes by 2026.

WM has also invested in emissions monitoring, acquiring Blue Sky Resources and the AirLogic analytics platform to further its efforts in landfill methane mitigation.

As Jim Fish, President and CEO of WM, stated,

“We are meeting our customers' needs today and investing in communities for the long term through our planned growth in recycling and renewable energy infrastructure.”

For deeper insights into WM's technological advancements and their regional impact, explore the details in San Antonio's coverage of the Mesquite Creek recycling facility launch, the comprehensive WM press release on their modernization strategy, and WM's official announcement via PR Newswire release on new recycling and renewable gas facilities.

Facility	Investment	Annual Capacity	Key Tech
Mesquite Creek (New Braunfels, TX)	$72M	144,000 tons	AI, 16 optical sorters
Total WM plan (2022-2026)	$3B	2.8M tons added	Automation, analytics

Emily Royall, Senior IT Manager for San Antonio's Emerging Technology Division, is redefining smart cities in Texas by championing responsible AI and data governance grounded in transparency, inclusion, and ethical innovation.

Royall's leadership has introduced community engagement programs such as the SmartSA Sandbox, where residents interact with digital tools, provide feedback, and participate in technology-driven services - attracting over 500 participants annually and collaborating with 25+ STEM organizations.

As a founding member of the GovAI Coalition, Royall ensures San Antonio joins a network of over 250 governments and 600 public servants dedicated to responsible, non-discriminatory AI use across public sector projects.

She emphasizes the operational need for transparency and oversight, stating,

“We promote a culture of experimentation and a willingness to navigate uncertainty...What my team has been building at the city of San Antonio is an operational process that gives the city a chance to ‘look under the hood' of new technologies and evaluate them before we make big commitments.”

San Antonio's inclusive, “people-centered” approach to smart city transformation addresses digital equity and public trust, with initiatives spanning rapid prototyping of AI solutions, collaborative policy development, and engaging vulnerable populations in digital skill building.

Royall's philosophy is clear: smart cities must ask who technology serves and measure outcomes for their communities - never adopting tech just for its own sake, as detailed in her recent Government Technology profile.

Metric	Value
Annual SmartSA Sandbox participants	500+ adults and children
STEM organizations collaborating	25+
GovAI Coalition governments	250+
Population represented by GovAI	150M+

A new AI-driven cyber threat, dubbed slopsquatting, is drawing concern across the software development and cybersecurity community. As generative code tools steadily reshape the industry, their well-documented habit of “hallucinating” convincing but non-existent software dependencies is being weaponized by threat actors; attackers register these phantom package names and publish malicious packages to repositories like PyPI and npm, opening the door to widespread supply chain attacks.

Recent large-scale studies by the University of Texas at San Antonio and partners reveal that roughly 20% of AI-recommended code packages did not exist, with hallucination rates much higher in open-source models (21.7%) than their commercial counterparts (5.2%).

Notably, CodeLlama models produced the most hallucinations, while GPT-4 Turbo performed best. The persistence of these hallucinations is especially troubling: 43% of hallucinated package names recurred in every trial, providing bad actors with a predictable pipeline to exploit.

Here's a quick data summary:

Model Category	Average Hallucination Rate
Open Source (e.g., CodeLlama, DeepSeek)	21.7%
Commercial (e.g., GPT-4 Turbo)	5.2%
Consistent Hallucinations Across Trials	43%

Developers are urged to cross-check every AI-suggested package before use, and to implement strict dependency and runtime validation tools.

As The Register's report on AI code suggestions and supply chain sabotage details, “slopsquatting” turns algorithmic fabrication into an attack vector as nefarious as traditional typosquatting, but far less reliant on human error.

According to research reported by CSO Online's coverage of AI hallucinations leading to slopsquatting threats, proactive measures such as model guardrails, dependency scanners, and developer education are key.

As Infosecurity Magazine summarizes,

“Developers trusting AI-generated outputs without proper validation increase risk exposure.”

For practical strategies and further findings, visit Kaspersky's detailed analysis of slopsquatting and prevention tactics.

San Antonio-based Wytec International has launched a 30-day "No-Cost" pilot for its advanced AI Gunshot and Drug Detection solution, targeting over 170 cities, 50 counties, and 40 special districts - including schools and hospitals - across the U.S. through the TXShare Cooperative Purchasing Program.

By integrating patented AI/ML algorithms, private LTE networks, and distributed sensor technology, Wytec aims to help municipalities and school districts address critical public safety threats with improved accuracy and real-time alerts.

The pilot leverages partnerships with Lemko Corporation and Nextivity Inc., deploying secure wireless infrastructure on FCC's CBRS spectrum, and achieves reported gunshot detection accuracy rates of 85%-95% in real-world educational settings.

The company - which has previously deployed its technology at the Johnson Space Center and major Dallas skyscrapers - has engineered the initiative to be scalable, with future expansion planned for more than 1,200 Texas ISDs and nationwide roll-out.

As Wytec hosts a technology seminar in San Antonio and eyes a Nasdaq uplisting, its collaborative, multi-patent approach promises a broad impact on public safety.

For more on this public safety innovation, visit the official Wytec Gunshot/Drug Sensing Sales announcement from BusinessWire, Yahoo Finance's feature on Wytec's AI pilot by Yahoo Finance, and a technical deep-dive from Private LTE & 5G News on Wytec's private network and sensor technology for Texas schools.

Scope	Pilot Program Coverage
US Cities	170+
Counties	50
Special Districts	40+

Wytec's AI-powered infrastructure "utilizes a private, secure wireless connection, eliminating potential interference during critical gunshot detection," ensuring rapid, reliable threat alerts for educational and municipal clients.

Generative AI's rapid adoption in San Antonio classrooms is sparking substantial debate among educators and experts, with concerns ranging from academic integrity to the future reputation of students and institutions.

Local and national surveys indicate that nearly 39% of students have used ChatGPT or similar AI tools for assignments, yet most colleges - including those in San Antonio - are still evolving their policies on AI use and detection.

As highlighted in The Chronicle's coverage on how AI is changing higher education, professors struggle to distinguish authentic student work from AI-generated content due to unreliable detection tools and the sophistication of new models.

A recent Forbes analysis reveals that, in just one year, Turnitin flagged over 22 million U.S. student papers with at least 20% AI-generated content, raising ethical and reputational risks for both students and schools, including threats of degree revocation and public scandals.

San Antonio educators are now urged to adopt transparent course policies, teach AI literacy, and redesign assignments for authenticity, as outlined in new district guidance.

A recent guide recommends,

“Do not rely solely on automated reports to determine academic dishonesty; prioritize assignment design.”

The evolving debate now considers not just policy enforcement, but also how to prepare students for careers where AI tools are integral.

See below for recent survey findings on student AI assignment use and institutional responses:

Survey/Source	Student AI Use	Institutional Response
Chronicle of Higher Ed	~25% of students use generative AI	Professors updating integrity policies; detection is difficult
Forbes/Turnitin Data	22M papers (20%+ AI); 6M (80%+ AI)	Mixed enforcement; some disable detection due to false positives
The Prairie News Survey	56% of college students used AI on assignments	Policy options: ban, regulate, or permit with citation

Explore more district guidance for faculty, including assignment redesign strategies and AI literacy resources, at SMCCCD Academic Senate's Plagiarism and Artificial Intelligence Guidance.

The recent Air Education and Training Command (AETC) summit in San Antonio spotlighted the Air Force's rapid integration of artificial intelligence (AI) to accelerate training, boost readiness, and drive ethical innovation across its ranks.

From launching immersive learning challenges utilizing cutting-edge platforms like Dreamscape Learn, to prioritizing robust data literacy and responsible, role-focused AI training spearheaded by the Air Force Research Laboratory (AFRL), AETC is reshaping how Airmen learn and evolve for autonomous roles.

This approach includes secure AI sandboxes and tailored guides for supervisors, HR staff, and administrative support, helping members navigate generative AI tools safely and strategically.

As AFCDA showcased its AI-driven curriculum engine, “The Robot,” which boasts up to 90% initial draft accuracy and dramatically shortens course development timelines, human oversight remains essential.

As Master Sgt. Adam Roush shared,

“We're at about a 90% accurate first draft. We may have to change some things, but it just goes to show the little things that we have to pick out and update, which makes it so rapid for us.”

Transparency also receives a boost from the new CLARA platform, centralizing progress, spending, and collaboration opportunities on AI initiatives.

For those interested in a deeper dive, explore the Force Development - Air Education and Training Command site for the latest newsletter, read about the Air Force's advances in AI-powered curriculum development at AFCDA, or get a comprehensive overview of how the new CLARA tool tracks Air Force AI progress.

This collaborative, ethical, and data-driven strategy places San Antonio at the forefront of AI-powered defense education.

San Antonio-based Wytec International is accelerating public safety innovation across the United States by deploying patented AI-enhanced gunshot detection and multi-threat sensor systems in over 170 cities, 50 counties, and 40+ special districts - including schools, police departments, and hospitals - under a major multi-state Master Services Agreement (Wytec AI Threat Detection Launch).

Wytec's proprietary solution combines 5G connectivity, in-building cellular, and AI/machine learning algorithms tested on hundreds of thousands of samples, yielding over 90% gunshot detection accuracy and significantly reducing false positives.

Their technology, developed in collaboration with Lemko Corporation, integrates advanced sensors detecting not only firearms but also drugs, chemicals, and other campus threats, with applications ranging from K-12 school districts to transit authorities.

According to Wytec CTO Robert Sanchez,

“The Lemko Corporation, with its patents on Distributed Mobile Architecture, will help solve America's growing issues of gun violence and drug epidemics affecting children and students in U.S. cities and Independent School District properties.”

Wytec, recognized as San Antonio's Best Tech Startup four times, is preparing to uplist to Nasdaq, a move set to unlock substantial funding for national expansion (Wytec Gunshot Detection for Schools).

The following table summarizes key aspects of Wytec's rollout and technology:

Deployment Scope	Core Technologies	Key Performance
170+ cities, 50 counties, 40+ districts	AI gunshot/drug sensors, Private LTE, in-building cellular	>90% detection accuracy; 2 US patents, 5 provisional patents

For more insight on Wytec's expansion and public safety tech leadership, explore their latest partnership and business updates at Wytec Nasdaq Uplisting Preparation.

Emily Royall, San Antonio's Smart City Administrator, has earned national recognition as one of the top 25 Tech "Doers, Dreamers & Drivers," reflecting her pivotal role in advancing ethical, people-centric innovation for the city's digital future.

With a background spanning urban planning, data governance, and tech policy - at both local and international levels - Royall leads the SmartSA initiative, a partnership connecting public utilities, transit, housing, and over 25 STEM organizations to ensure technology-driven improvements equitably benefit San Antonio's 1.5 million residents.

Under her leadership, accessible events like SmartSA Sandbox have empowered more than 500 adults and children annually to engage with hands-on technologies and offer feedback for city pilots, helping San Antonio scale 10-15 smart city prototypes each year.

For Royall, collaboration and transparency are essential:

"I believe people should be at the center of smart cities, and work across sectors to enhance public oversight of smart city technologies and the data that powers them."

Her approach directly addresses pressing community challenges - San Antonio has the highest poverty rate among major U.S. cities - and seeks to build enduring public trust through inclusive engagement and coalition-building, such as participation in the GovAI initiative representing over 150 million people nationwide.

To learn more about Royall's vision for transparency, accountability, and community involvement in urban innovation, see her profile on the San Antonio Smart Cities Team page, her perspectives in Redefining Smart: Transform Magazine, and discover how SmartSA Sandbox brings residents and technology together in Smart Cities Dive's coverage.

This month, a groundbreaking study by UTSA and collaborators reveals systemic security risks posed by AI-generated code tools, with "package hallucinations" - where AI models suggest non-existent software packages - emerging as a new threat.

Researchers analyzed 16 code-generating large language models (LLMs) and found that nearly 20% of recommended packages were hallucinated, with open-source models like CodeLlama showing a 21.7% rate compared to just 5.2% in commercial models like GPT-4 Turbo.

These hallucinations enable a rising cyberattack technique dubbed slopsquatting, where attackers quickly register fake packages with convincing names, making it easy for malicious code to be distributed through trusted AI-generated recommendations.

The table below summarizes key findings:

Model Type	Hallucination Rate	Worst Offender	Total Tested Packages
Open Source	21.7%	CodeLlama (>33%)	2.23 million
Commercial	5.2%	GPT-4 Turbo (3.59%)	2.23 million

“Package hallucinations were found to be a systemic issue across all models and languages tested, including state-of-the-art commercial models... A key component of a package hallucination attack... is that they are extremely easy for a threat actor to produce with essentially zero cost or risk.” - Joe Spracklen, UTSA researcher

San Antonio's leadership in disclosing these vulnerabilities has prompted major AI vendors to respond, and experts urge developers to use dependency scanners and cross-reference package names, while also calling for robust improvements in LLM development.

For a detailed breakdown of the study and the emergence of slopsquatting as a supply chain threat, read the original UTSA research highlights on AI-generated code security risks, a comprehensive industry analysis of AI code tool hallucinations, and an in-depth overview of AI hallucinations creating new software supply chain threats.

San Antonio stands at a pivotal nexus where rapid technological advancement meets the growing demands of ethical responsibility. The city's innovation scene reflects both the surge of AI-driven startups - mirroring national trends, as U.S. AI companies like Anthropic and OpenAI reach multibillion-dollar valuations and together draw over $40 billion in investment this year alone (19 AI startups raising $100M+ in 2025) - and intensifying efforts to ensure technology is developed and deployed responsibly.

Texas legislators have responded with decisive action, passing House Bill 366 to mandate transparency in political ads that use altered media or generative AI, underscoring concerns over misinformation and digital ethics.

As Rep. Dade Phelan noted,

“This is the beginning of a new era in ethics where the voters need to know what is real and what is not.”

The bill, which now moves to the Senate, aims to safeguard voters amid the growing sophistication of AI-generated content (Texas House AI transparency bill details).

Meanwhile, global initiatives like the United Nations' AI for Good Innovation Factory illustrate how San Antonio's tech startups can leverage mentorship and funding while adhering to criteria for innovation, market fit, and ethical risk management (AI for Good Innovation Factory overview).

As the city advances, fostering local AI talent through accessible education - such as Nucamp's range of bootcamps and scholarships - will be essential in shaping a workforce ready to navigate not only rapid employment shifts but also the ethical frontiers defining tomorrow's technology.