General February 2, 2026 Our News RoomBy Hema Dey
Read the full article here:
Two recent pieces have stayed with me.
One is Meredith Whittaker’s, CEO of Signal warning that “encryption is deeply threatening to power.” The other is Dario Amodei’s, Anthropic’s co-founder framing of AI as a technology in its adolescence — powerful, transformative, and dangerously ahead of our collective maturity.
On the surface, they address different domains: privacy and messaging on one hand, artificial intelligence on the other. But together, they point to the same uncomfortable reality:
We are redistributing power faster than we are willing to accept responsibility for it.
And regulation is not keeping up.
Watching Meredith Whittaker’s recent interview with Fareed Zakaria on his regular Sunday show on CNN, what struck me most wasn’t ideology. It was urgency.
You could feel the shock and horror she experienced when messages associated with Pete Hegseth, the US Secretary of War, were reported as leaking through Signal instantly turning a personal or operational failure into a referendum on Signal itself. What followed was a brutal reputation-management triage: painful, public, and relentless. And yet, ultimately, it was successful.
That moment matters because it reveals something we often gloss over:
Even the most privacy-first, principled systems are stress-tested not by theory, but by power, behavior, and human error.
Encryption didn’t fail because it was flawed. It was tested by real-world complexity.
Amodei’s metaphor of AI’s adolescence is apt. Adolescence is a stage of explosive capability, creativity, and confidence — paired with immaturity, impulsiveness, and blind spots.
That’s exactly where we are with AI.
Governments are behind. Legislation is fragmented, reactive, and often written by people who do not fully understand the systems they are attempting to regulate. AI governance today feels more like early aviation than modern infrastructure: breathtaking progress, limited guardrails, and rules written only after something breaks.
We are, collectively, learning on the go.
But that doesn’t disperse responsibility. It concentrates it.
Whittaker’s argument about encryption applies almost perfectly to AI.
Encryption threatens centralized surveillance.
AI threatens centralized expertise.
Both redistribute power — away from institutions and toward individuals.
And power never disappears. It moves.
The real question isn’t whether these technologies are dangerous. It’s who bears responsibility when things go wrong.
Until regulation meaningfully catches up, the answer is simple and uncomfortable:
The buck stops with individuals — builders, deployers, executives, founders, and yes, billionaires.
Waiting for regulation before acting ethically is not neutral. It’s a choice.
Here’s the part we don’t like to say out loud.
For AI to be genuinely useful, information must be shared.
Whether it’s:
AI cannot meet our needs in a vacuum.
Context is oxygen.
In my world marketing, automation, and agentic systems usefulness depends on access to intent, history, preferences, and performance data. Sometimes that data is sensitive. Often it’s deeply personal.
The goal, then, is not zero data sharing. That’s a fantasy.
The goal is intentional, bounded, and ethical data sharing.
This is where encryption and AI intersect most clearly.
Privacy isn’t about pretending data doesn’t exist. It’s about who controls it, who benefits from it, and who bears the risk when something goes wrong.
Today, many AI systems:
When failures occur, users are left holding the consequences — reputationally, financially, or emotionally.
That’s why moments like Signal’s crisis feel so visceral. They expose how thin the margin for error really is.
Agentic AI doesn’t just analyze. It acts.
It responds, books, routes, decides, and escalates often faster than humans can intervene. That speed is powerful, but it also means mistakes compound rapidly.
Nowhere is this clearer than in regulated environments like healthcare.
On paper, HIPAA is robust:
In practice, AI exposes the gaps.
LLMs are probabilistic, not deterministic.
Voice transcription pipelines leak metadata even when transcripts are encrypted.
Human-in-the-loop review slows response times — but removing it increases risk.
So where does practicality sit?
Not in pretending compliance alone is sufficient.
It sits in design discipline:
Compliance may satisfy an audit.
Ethical design protects people.
One of the biggest myths in AI is that safety slows innovation.
What actually slows innovation is:
The teams that move fastest long-term are the ones that design restraint into the system from the beginning.
Despite the risks, I remain optimistic.
AI is not just a risk amplifier. It is also one of the most powerful equalizers we’ve ever seen.
It lowers barriers to expertise.
It compresses learning curves.
It enables scale without headcount.
But this only holds true for those who begin their journey now.
The real divide won’t be human versus machine. It will be AI-literate versus AI-illiterate.
That’s why access, education, and ethical deployment must grow together — or power simply recentralizes in fewer hands.
Technological adolescence is not a reason to slow to a halt. It’s a reason to grow up quickly.
AI will not mature on its own.
Encryption will not defend itself.
Ethics will not emerge by accident.
Until legislation catches up — and it will — individual accountability remains our most important guardrail.
Every system we deploy reflects a choice:
The danger is not that we are moving fast.
The danger is pretending we are not responsible while we do.
If you want to chat about how to take ownership of your conduct, your contribution to safety, guardrails, systems and security – I would love to hear your thoughts.
Yes — and that’s exactly the point.
Right now, there is no external system sophisticated enough to meaningfully govern fast-moving, agentic AI. Waiting for perfect regulation is a comforting idea, but it’s not realistic. In this gap, values become architecture.
Every builder decides:
How much data is truly necessary
Whether systems remember by default
How errors are handled
Who can intervene and when
Those decisions shape outcomes more than any policy document. This isn’t about blind trust — it’s about owning responsibility instead of outsourcing it to future legislation.
Only if it’s done carelessly.
AI can’t be useful without context — that’s the uncomfortable truth. But privacy isn’t about refusing to share data; it’s about control, scope, and intent.
Ethical systems ask:
What data is needed for this task — and nothing more?
How long does it live?
Who can access it?
What happens when something goes wrong?
The danger isn’t AI knowing things.
The danger is AI knowing things indefinitely, unnecessarily, and without accountability.
I don’t think so — I think we’re finally being honest.
The idea that safety slows innovation is outdated. What actually slows innovation is:
Rebuilding after breaches
Losing user trust
Regulatory overcorrections triggered by avoidable harm
The teams that move fastest over time are the ones that design restraint early. Ethical guardrails don’t kill momentum — they protect it.
AI is an extraordinary equalizer, but only if we engage with it deliberately. Speed without responsibility isn’t progress. It’s a liability.
Call Us: 949-779-6442
If you are falling in love with us,
learn our love language with this eBook
before we seal the deal...
