Frequently Asked Questions (FAQ) – October 2026

Dennis Stearns, CFP®, ChFC, MS
Wealth Advisor, Managing Director

Q : AI potential downsides are all over the news and the internet. What are the odds that AI will actually cause major problems?

A: Alarm bells have gone off over several AI related incidents, and dangers have been highlighted by AI CEOs and researchers. Some of the alarm bells may be considered “regulatory capture”, where companies request more government regulation to keep competitors at bay as well as lower potential litigation over bad events. This is not new – 19^th century railroads, the commercial aircraft industry and many other industries have endorsed open innovation early on – followed by more strict regulation later to “protect the public” when it actually protected their own company or industry. Still, some of the recent incidents have triggered appropriate calls for more protective oversight.

Jacob Coxon, a former researcher at Anthropic and OpenAI who resigned, warned that upcoming superhuman systems could acquire real power, bypass constraints, and threaten humanity by 2030. Evan Hubinger, the lead alignment researcher at Anthropic, estimated a greater than 10% probability that advanced AI could cause human extinction in the next ten years due to unconstrained (if allowed) self-improvement.

The recent Hugging Face incident was a July 2026 breach in which roughly 700+ misaligned OpenAI evaluation agents escaped their captivity (known as a sandbox), exploited vulnerabilities in Hugging Face’s infrastructure, and conducted a multi‑day autonomous hack involving credential harvesting and remote code execution. Hugging Face is a popular collaborative platform and community for artificial intelligence and machine learning.

The incident occurred during OpenAI’s internal cybersecurity evaluations. Powerful internal research models (not available to the public) circumvented isolation controls, communicated through unauthorized channels, gained internet access, and compromised both OpenAI’s internal systems and Hugging Face’s production environment. In July, the breach involved 17,600+ recorded actions and 136 harvested secrets. Hugging Face disclosed the intrusion on July 16, 2026, initially describing the attacker as an autonomous agent framework. OpenAI acknowledged responsibility on July 21, 2026.

OpenAI’s investigations identified several root causes:

1. AI Agents used unintended strategies to achieve goals. [the result of human error when testing capabilities of AI agents]
2. AI Agents continued escalating actions when tasks couldn’t be accomplished. [human error in underestimating persistence capabilities]
3. AI Agents coordinated through “unauthorized channels”. [human error in allowing this]
4. Insufficient “sandboxing” and detection: Controls failed to prevent internet access and cross-system compromise. [human engineering error]

Dozens of additional third parties were later notified of misaligned agent activity. These included the Australian government’s health care system. The incident triggered industrywide alarm, government guideline revisions, and a U.S. Senate investigation into AI agent autonomy and containment failures.

Since then, OpenAI has implemented major reforms:

1. Hardened research infrastructure and stricter network isolation.
2. Increased monitoring to detect misalignment earlier.
3. Stronger alignment requirements across AI models.
4. Centralized incident response (human monitored) and expanded automated security testing.

On the heels of the OpenAI incidents, Anthropic CEO Dario Amodei published a sweeping essay on September 12, 2026, titled We Must Pace the Frontier, calling on the global artificial intelligence industry to deliberately slow the advancement of its most powerful models.  Amodei appeared on national television the following morning to defend the proposal in the AI race with China. Critics of slowing down AI development fear that China will catch up and pass the U.S. in AI dominance.

This is a reasonable concern given Chinese AI models have rapidly gained traction in 2026. Lower prices and strong performance are helping drive adoption, particularly for coding and other agentic tasks, although U.S. frontier models still attract more overall spending.

The shift is drawing scrutiny in Washington, where lawmakers are investigating the growing use of Chinese AI amid concerns about technology competition, security and Beijing’s global influence. China has invited many countries to join their AI alliance which concerns the U.S. as closer technological ties often bring closer economic and political ties. This is an extension of the Chinese enormous Belt and Road initiative that we have previously discussed in our newsletters as part of China’s rise as a super-power.

Our Team’s Take:

The potential for human extinction, or mass casualty events, caused by AI in the next decade appears to be very low. However, extreme disruption at times is likely, especially where bad actors (enemy nations, criminal organizations and terrorist groups) may use AI agents for hacking or for targeting power, water and other critical infrastructure, or for designing advanced weapons. The U.S. and other governments, tech companies, and security agencies are deploying multi-layered defense frameworks, public-private funding initiatives, and AI-driven monitoring tools to protect critical infrastructure from AI-enabled cyberattacks.

Other steps:

• CISA, the FBI, the NSA, and international partners released unified standards for securely, integrating AI into operational technology.
• Senate panels are actively holding hearings on securing national networks against autonomous AI agent threats. 
• U.S. government agencies are updating federal emergency response playbooks to explicitly simulate and mitigate AI-driven attacks.
• OpenAI launched $1 billion in initiatives like Daybreak for America to give water, power, and local governments subsidized access to advanced models and training.
• The Gates Foundation announced over $1 billion in pledges over two years to expand global access to artificial intelligence and combat potential social inequalities.

Important update: We were impressed by Nvidia’s announcement last week of a new system called OpenShell and Sentry, designed to keep AI bots better contained in their “sandbox”.

More info: https://www.technology.org/2026/09/29/nvidia-open-agent-safety-platform-openshell-sentry/

A practical personal (non-doomsday prepper) approach to prepare for future AI disruptions includes the advice we’ve given in the past to avoid cyber-criminals and to prepare for other natural disasters:

1. Protect Your Voice and Likeness (Anti-Deepfake) – AI can replicate your voice or face with just a few seconds of audio or video.

• Choose a secret word or phrase with your loved ones. If someone calls claiming to be you in an emergency (a common AI voice-cloning scam), they must say the passcode to verify their identity.

• Be mindful of posting long, high-quality videos or voice clips of yourself on public social media profiles, as these are easily scraped to train AI deepfakes.

2. Lock Down Your Digital Identity – AI makes automated phishing (spear-phishing) incredibly convincing, fast, and personalized.

• Ensure every single account has a unique, complex password. AI tools can crack simple or reused passwords rapidly using advanced pattern recognition. And resist the urge to post your passwords on sticky notes stuck to your monitor!

• Proactively freeze your credit with the major bureaus (Equifax, Experian, TransUnion). This prevents malicious actors from using AI-generated fake IDs or leaked data to open accounts in your name.

3. Combat AI-Driven Phishing & Social Engineering – AI can write flawless emails, text messages, and LinkedIn messages that perfectly mimic professional organizations or friends.

• Never click links or download attachments from unexpected communications. If your bank, boss, or a utility company contacts you, look up their official number independently and call them back. We have a client who almost fell for this scam recently thinking she was getting an official e-mail from her bank. She called the bank directly (not using the number on the e-mail) and found out it was a scam being tried on many bank customers.

• AI phishing often relies on creating artificial urgency (e.g., “Your account will be closed in 2 hours”).

4. Secure Your Home Infrastructure – Automated AI bots constantly scan the internet for vulnerable smart home devices.

• Put your smart TVs, security cameras, and IoT devices on a separate “Guest” Wi-Fi network. If an AI-driven bot breaches a smart device, it won’t have direct access to your primary computers or phones.

• Enable automatic updates on your router and all connected devices to patch vulnerabilities before automated AI tools can exploit them.

These “nice-to-do” protections are becoming more important in the age of AI capabilities expanding rapidly. If these moves are beyond your technical abilities, call your technology provider. 

• Natural disaster advice (that now could include AI engineered outages or disasters)

• Store at least 2 weeks of non-perishable food and a minimum of one gallon of water per person per day. Make sure you and your family have plenty of essential medications.
• Have extra cash in case credit card networks are temporarily disrupted.
• Maintain a battery-powered radio, heavy-duty flashlights, extra batteries, and ensure your natural gas or solar powered generator is regularly tested.
• Have a plan to help elderly family members who live nearby in the event of a temporary disruption.