technology

Deepfakes Kill… Make them criminal!

cantechit Uncategorized , , , ,

It’s time to talk about deepfake technology. You know, those AI tricks that fabricate videos or audio making it seem like someone said or did stuff they never did? It’s blurring reality so badly, you have to ask: “What if that viral clip of a celeb or politician tomorrow is total BS?” Or scarier, what if it’s targeting you or your family?

This tech is evolving at breakneck speed, and it’s shockingly easy to misuse. What used to need pro-level gear and $100K in GPU now happens in minutes with free apps and AI models. Deepfake content jumped from about 500,000 files in 2023 to a projected 8 million in 2025. The number of deepfakes detected globally across all industries increased 10x in 2023 (eftsure.com). Open-source tools and cheap computing mean anyone—hackers, trolls, or kids—can harness it. This is a recipe for problems.

Misuse is rampant, starting with fraud that’s hammering Canadians. We’ve lost $103 million to deepfake scams in 2025 (Mitchell Dubros), with North American cases up 1,740%. And 95% of Canadian companies say deepfakes have increased their fraud risk. Deepfakes now account for 6.5% of all fraud attacks, marking a 2,137% increase since 2022. One example? A firm lost $25 million to a deepfake CEO scam. Would you catch a deepfaked loved one asking for cash?

But the real horror is – none of this is directly a criminal offence.

Harassment, especially against young people, leading to tragedy. Deepfakes fuel bullying, extortion, and non-consensual porn—mostly targeting women and girls. In Canada, a pharmacist was linked to the world’s most notorious deepfake porn site, and Alberta cops warn of kids sharing AI fakes. Receipts: A Faridabad student died by suicide over deepfake blackmail. Twitch streamer QTCinderella (Blaire) faced humiliation in 2023, and Q1 2025 saw 179 incidents, up 19% from all of 2024 (keepnetlabs.com). Sextortion using deepfakes has driven suicides amid blackmail and isolation. If you’re a parent, think: How protected are our kids from this digital nightmare? There is no protection for this conduct under law – at least not directly.

C-63 in Canada targets PLATFORM OPERATORS – it stops short of the important step of making non-consensual deepfakes of others a crime. If someone deepfakes your kid, and plasters it on TikTok – it’s too late, the harm is done. We need to prevent this in the first place with severe consequences under law.

Deepfakes erode trust—in elections (like those fake videos of Canadian politicians), relationships, everything. Youth, always online, suffer most from amplified cyberbullying that can end in depression or worse. Europol warns deepfakes fuel harassment, extortion, and fraud, and detection lags behind the tech.

Call To Action

My plea: Governments must legislate now to shield the public, especially young folks. Canada is lagging behind the US on this. We’ve got piecemeal stuff like the Elections Act on campaign fakes and Bill C-63’s start on harms, but no full law tackles everyday non-consensual deepfakes. We need it classified as a serious criminal offense under the Criminal Code of Canada—not a slap on the wrist, but with hefty fines and jail time for creating deepfakes of anyone without consent.

I am calling on our Canadian Government to enact changes to the Criminal Code of Canada to specifically make non-consensual deepfakes a serious criminal offence, with stiff fines and jail time.

Act now: Contact your MP and demand this change. Share this, educate your circle, teach kids to spot fakes. Platforms must remove deepfakes fast, schools need education programs. Why risk more lives? Let’s make deepfakes a crime that bites back—before it’s too late. What are you waiting for?


My Proposed Legislation

Non-consensual deepfakes

162.3 (1) In this section,

deepfake means a visual or audio recording that is created or altered using artificial intelligence or other technology in a manner that would cause a reasonable person to believe it depicts the person engaging in conduct or speech that did not occur, and includes any synthetic representation that is realistic in appearance or sound;

distribute includes to transmit, sell, advertise, make available or possess for the purpose of distribution.

(2) Everyone commits an offence who, without the express consent of the person depicted, knowingly creates, distributes or possesses a deepfake of that person if

(a) the deepfake depicts the person in an intimate context, including nudity, exposure of genitals, or explicit sexual activity; or

(b) the creation or distribution is intended to cause harm, including emotional distress, reputational damage, or incitement to violence against the person.

(3) For the purposes of subsection (2), consent must be informed, voluntary and specific to the creation and use of the deepfake, and may be withdrawn at any time; however, no consent is obtained where the agreement is obtained through abuse of trust or power, or where the person is incapable of consenting.

(4) Subsection (2) does not apply to

(a) deepfakes created with informed consent by the depicted persons that has not been specifically revoked.

Punishment

(5) Everyone who commits an offence under subsection (2) is guilty of

(a) an indictable offence and liable to imprisonment for a term of not more than five years; or

(b) an offence punishable on summary conviction and liable to a minimum fine of $1000 but not more than $25,000 or to imprisonment for a term of not more than two years less a day, or to both.

(6) In determining the sentence, the court shall consider as aggravating factors

(a) whether the offence involved a minor or vulnerable person;

(b) the extent of harm caused to the victim; and

(c) whether the offender profited from the offence.

The Edge, Reimagined: Why Cisco Unified Edge is the Mind-Shift We Needed

cantechit Uncategorized , , , ,

Let’s face it: the edge has long been a “necessary evil.” Nobody wants fat servers, complex infrastructure, and constant management headaches in remote locations, but it’s been unavoidable. The cloud, while powerful, can’t solve everything, especially when it comes to the low-latency, GPU-intensive demands of modern AI, or the pervasive issue of vendor lock-in. My contention? This old way of thinking about the edge is over.

If the original Raspberry Pi was that plucky, credit-card-sized marvel that let hobbyists and tinkerers dream up all sorts of clever, small-scale computing projects, then the Cisco Unified Edge is like its ridiculously buff, impeccably dressed, and highly intelligent older sibling who just graduated from a top-tier business school with a PhD in AI.

Cisco’s new Unified Edge isn’t just another product; it’s a total mind change. We want less at the edge – less complexity, less hardware – but more power where it counts. AI needs GPUs and low latency, and you can’t always get that efficiently from the cloud.

This platform addresses that head-on. It’s an integrated, modular system combining compute, networking, storage, and security, purpose-built for distributed AI workloads. It brings the necessary power, including GPU support, right to the source of data generation. Think real-time AI inferencing on a factory floor or in a retail store, without the latency penalty of sending data halfway across the globe.

Crucially, it’s not the “same old architecture.” Cisco Unified Edge simplifies operations with features like zero-touch deployment and centralized management via Cisco Intersight, transforming the edge from a burden to a strategic asset. Security is baked in, addressing the expanded attack surface of distributed environments.

This isn’t just about putting more powerful chips at the edge; it’s about a fundamental architectural shift at the edge, driven by the integrated power of a System on a Chip (SoC). Instead of separate, bulky components for compute, networking, and security, Cisco Unified Edge leverages Intel Xeon 6 SoC processors. This level of integration is the game-changer, allowing for a far more compact, efficient, and unified platform that delivers the necessary AI-ready performance, including GPU support, without the traditional sprawl and complexity. It’s how Cisco achieves “less at the edge” in terms of physical footprint and management overhead, while simultaneously providing “more power” right where real-time AI inferencing and agentic workloads need it most, truly transforming the edge from a patchwork of devices into a cohesive, intelligent brain.

As Cisco’s Jeetu Patel noted, “Today’s infrastructure can’t meet the demands of powering AI at scale”. Cisco Unified Edge changes that. It provides the raw compute and GPU muscle for demanding AI at the edge, but in a lean, intelligent, and manageable way. It transforms the edge from a reluctant necessity into a strategic advantage, allowing sophisticated capabilities to flourish where they’re needed most.

This is a different way of thinking at the edge, and I like it. A lot. It’s going to change the game.

Agentic AI vs Deterministic Code

cantechit Uncategorized , , , ,

No question – Building apps with LLMs in agentic setups is a game-changer, but it can also be a pain in the butt compared to good old deterministic code. Craft an clever agent that summarizes docs or fixes bugs, then bam, the model updates, and suddenly it’s spouting nonsense, ignoring prompts or ignoring basic words like “yes”. Non-deterministic chaos at its finest.

Deterministic code? It’s the reliable workhorse: feed it input X, get output Y every damn time. Fixed rules, easy debugging, perfect for stuff like financial calcs or automation scripts where surprises mean lawsuits. As Kubiya nails it, “same input, same output”—no drama.

“A computer will do what you tell it to do, but that may be much different from what you had in mind”.  – Joseph Weizenbaum — Not when your using a model you probably didn’t build and not your own weights.

Agentic AI with LLMs? That’s the wildcard party crasher. These systems think on their feet: reason, plan, grab tools, adapt to goals like tweaking marketing on the fly or monitoring health data. IBM calls it “agency” for a reason—it’s autonomous, pulling from real-time vibes beyond rigid training. But here’s the kick: it’s probabilistic. Outputs wiggle based on sampling, context, or those sneaky model tweaks from OpenAI or whoever. LinkedIn rants about it: “Same prompt, different outputs.” Your app morphs overnight, and fixing it? Good luck tracing probabilistic ghosts.

This shift sucks for dev life. Traditional code: bug? Trace, patch, done. Agentic? Hallucinations, inconsistencies, testing nightmares. Martin Fowler compares LLMs to flaky juniors who lie about tests passing. It’s a paradigm flip—from control to “let’s see what happens.” Salesforce says pick deterministic for regulated certainty, agentic for creative flex. But non-determinism could means security holes, data risks, and endless babysitting. It also adds this attack vector that is really non-deterministic if the model has access to data that it needs to work – but I might not want exposed.

AspectDeterministic CodeAgentic AI with LLMs
PredictabilityRock-solid: Always consistentSketchy: Varies like the weather
AdaptabilityStuck to your rulesBoss: Handles dynamic crap
Testing/FixingSimple: Logic checks and patchesHell: Variability demands tricks
Best ForPrecision gigs (finance, compliance)Goal-chasing (support, optimization)
Pain LevelLow: Set it and forget itHigh: Constant surprises

Bottom line: Hybrids are the way—LLMs for the smarts, deterministic for the reins. Deepset pushes that spectrum view: not binary, blend ’em. It sparks innovation, sure, but don’t romanticize—the annoyance is real. Code with eyes open, or get burned. Put humans in the loop to keep things in check.

What about Agentic AI ops for network and technology? Didn’t we just say “precision gigs” are better with deterministic code? That won’t stop the likes of awesome developers like John Capobianco https://x.com/John_Capobianco from pushing those limits, and he has been doing that for years at this point. Handing AI agents the keys to critical stuff like network monitoring, anomaly detection, or auto-fixing outages. Sounds efficient, right? But it’s a powder keg from a security standpoint. These autonomous bad boys can hallucinate threats, expose data, or open doors for hackers through memory poisoning, tool misuse, or privilege escalation. Cisco nails the danger: “The shift from deterministic code to probabilistic chaos is at the heart of securing AI agents that think for themselves,” highlighting a “lethal trifecta” of data leaks, wild hallucinations, and infrastructure weak spots that could cascade into total meltdowns.

Tool are starting to emerging though for AI security, particularly from Cisco and open-source communities to advance defenses against threats like prompt injections and supply chain attacks, but there is work to be done. Things like Cisco’s open-source Foundation-sec-8B model, a specialized LLM for cybersecurity tasks such as threat intelligence and incident response, will help developers start to build customizable tools with on-prem deployments to reduce hallucinations and enhance SOC efficiency. Their Hugging Face partnership bolsters supply chain security with an upgraded ClamAV scanner detecting malware in AI files like .pt and .pkl. Broader open-source efforts include Beelzebub for malicious agent analysis and Promptfoo for LLM red-teaming, yet challenges from hackers with evolving adversarial tactics using LLM’s to attack LLM’s are very much a thing…. The system is attacking the system being protected by the system… Yeah that.

Cisco-Hugging Face ClamAV Integration: https://blogs.cisco.com/security/ciscos-foundation-ai-advances-ai-supply-chain-security-with-hugging-face
Cisco Foundation-sec-8B: https://blogs.cisco.com/security/foundation-sec-cisco-foundation-ai-first-open-source-security-model

So much more to learn, but with all of that said…. Humans in the loop is going to be a thing for awhile – at least until Skynet…

Cisco Live 2025: Community, Balance, and Big Dreams for AI

cantechit Uncategorized , , , ,

That was huge.

Still buzzing from Cisco Live 2025 in San Diego. This wasn’t just a tech conference—it was a reunion of brilliant minds and big hearts. The Cisco Community Champions dropped wisdom that flipped my perspective, like a breakout session chat that rewired how I think about collaboration. Tech Field Day delegates brought the heat over late-night tacos, debating tech’s future with ideas that stuck with me. And my Cisco colleagues and friends? They’re family—coffee in the DevNet Zone, laughs at the Customer Appreciation Event (The Killers absolutely slayed!), and moments that recharge my soul. But as I look ahead, I’m thinking about balance, mentorship, and how we’ll make a real difference with AI. Here’s the vibe.

The Community That Fuels Us

Cisco Live is all about connection. Those conversations with Champions, delegates, and friends aren’t just chats—they’re sparks that ignite new ideas. A mentor’s advice over drinks is already shaping my next move, and the energy at the CAE was pure magic. This community pushes us to dream bigger and work smarter, together. Those that challenge me, thank you for doing that. The support from those who love our work challenges us to do more. Without community – we really have nothing.

What’s Next for Me: Building, Mentoring, and Balance

This year, I’m all about building. I’m diving into relationship building, leveling up my skills in innovative problem-solving, and finding new ways to share with you all. I’m hyped to get back to blogging, maybe even start vlogging, but I’m keeping it real—it’s a lot. Tech can be a grind, and we don’t always talk about the psychological toll it takes. The pressure to stay ahead, the endless hustle—it weighs on us. I’m prioritizing balance, making time for myself, and I invite you to do the same. Check in on your friends and colleagues, be that supportive ear. We’re stronger when we lift each other up.

I’m also building a structured mentorship plan to guide others, inspired by my own mentors. Whether it’s sharing tech insights or navigating career challenges, I want to pay it forward and help others shine. Who knew my greatest challenge by my own mentors, would be to pay it forward. I have started to realize and understand that climbing this career mountain hits a plateau, and unless you can lead a team to the top – you are now stuck.

Making a Difference with AI and Country Digital Acceleration

This year, I’m wrestling with a big question: How will I make a meaningful difference with AI? It’s consuming my thoughts. AI has so much marketing hype – I want to get past that. AI has the power to transform lives—think smarter cities, safer communities, inclusive access to tech – or just making super complex things – easier. At Cisco Innovation Labs, we’re celebrating 10 years and the anniversary of Country Digital Acceleration (CDA). I’m so grateful for CDA’s support, backing projects like Digital Canopy that bring connectivity and hope to underserved areas. Their belief in our ideas fuels us, and I’m stoked to deepen our work together, dreaming up solutions that change the world. This is a great partnership, and it really gives us the ability to “Design with Empthy, Innovate with Purpose”

The Next Big Thing for Our Labs

With a decade in the rearview, it’s time to go big. What’s the next big thing for Cisco Innovation Labs? I’m obsessed with figuring this out. Maybe it’s AI-driven public safety tools, or … well… so many things I can’t talk about yet… , or sustainable tech that powers a greener future. Whatever it is, it’ll be bold, human-centered, and built with this incredible community. I’m ready to dream, experiment, and make waves. I know one thing, technology comes second, people, community and EMPTHY come first.

Keep the Vibe Going

Cisco Live 2025 was a love letter to community, a reminder to stay connected and take care of ourselves. As I chase big dreams with AI and our Labs, I’m carrying this energy forward. So, take a moment for you, check in on your people, and let’s dream big together. What’s your Cisco Live highlight? Hit me up on Twitter or drop a comment—let’s keep it rolling!