WEBVTT 00:00:00.000 --> 00:00:02.879 Imagine for a moment you're sharing an AI conversation 00:00:02.879 --> 00:00:05.120 with a colleague. Maybe it's a brilliant new 00:00:05.120 --> 00:00:08.560 marketing idea or even a sensitive project proposal. 00:00:09.220 --> 00:00:11.419 What if that whole discussion full of private 00:00:11.419 --> 00:00:14.759 details just suddenly popped up on Google? For 00:00:14.759 --> 00:00:17.280 anyone to see, this isn't some far -fetched fear. 00:00:17.620 --> 00:00:19.579 It was actually a very real thing for a lot of 00:00:19.579 --> 00:00:21.800 people just recently. Yeah, it was a proper wake 00:00:21.800 --> 00:00:23.879 -up call, wasn't it? Kind of like finding your 00:00:23.879 --> 00:00:27.460 private diary. just displayed in the town square, 00:00:27.879 --> 00:00:30.460 a true stark reminder of our digital footprint 00:00:30.460 --> 00:00:33.539 these days. Welcome to the Deep Dive. Today we're 00:00:33.539 --> 00:00:35.439 digging into a pretty critical incident from 00:00:35.439 --> 00:00:40.320 mid -2025. It was when a technical glitch exposed 00:00:40.320 --> 00:00:42.500 private chat GPT conversations to the public 00:00:42.500 --> 00:00:45.340 internet. But this deep dive isn't just about 00:00:45.340 --> 00:00:47.859 that one platform, right? It's really a bigger 00:00:47.859 --> 00:00:50.679 lesson in digital privacy for all of us. Exactly. 00:00:51.320 --> 00:00:53.420 Our mission today is, well, first to help you 00:00:53.420 --> 00:00:55.899 understand what actually happened, then guide 00:00:55.899 --> 00:00:58.439 you through the immediate steps to clean up any 00:00:58.439 --> 00:01:01.140 potential past exposures from that. After that, 00:01:01.299 --> 00:01:03.630 we'll kind of zoom out. look at other AI data 00:01:03.630 --> 00:01:06.090 risks lurking around. Then we'll talk about building 00:01:06.090 --> 00:01:08.609 safer habits, definitely. And we'll even show 00:01:08.609 --> 00:01:11.290 you a smarter, much more secure way to share 00:01:11.290 --> 00:01:14.049 AI -generated stuff. And finally, yeah, we'll 00:01:14.049 --> 00:01:16.090 unpack this really sophisticated prompt that 00:01:16.090 --> 00:01:18.849 basically acts like your own personal AI safety 00:01:18.849 --> 00:01:21.150 co -pilot. OK, so it's all about understanding 00:01:21.150 --> 00:01:24.489 what went wrong, how to react now. and crucially, 00:01:24.810 --> 00:01:27.129 how to build a much safer workflow going forward 00:01:27.129 --> 00:01:30.090 in this AI world. Let's unpack this. So this 00:01:30.090 --> 00:01:32.349 specific incident with ChatGPT where the shared 00:01:32.349 --> 00:01:34.650 links somehow got indexed by search engines, 00:01:35.170 --> 00:01:37.069 it really exposed a kind of hidden vulnerability. 00:01:37.390 --> 00:01:39.370 Now, OpenAI, they acted quickly to fix it, which 00:01:39.370 --> 00:01:42.349 is good. But even with that fix, some data, like 00:01:42.349 --> 00:01:44.269 a digital residue, might still be lingering out 00:01:44.269 --> 00:01:46.250 there, just floating around. That's exactly right, 00:01:46.290 --> 00:01:48.609 like a digital ghost in the machine, as you said. 00:01:49.090 --> 00:01:52.019 So the first... And honestly, the most fundamental 00:01:52.019 --> 00:01:54.420 step, if you've ever created a shared link through 00:01:54.420 --> 00:01:57.420 ChatGPT, is simply to delete it right from inside 00:01:57.420 --> 00:01:59.939 the platform. Go log into your account, click 00:01:59.939 --> 00:02:02.180 on your profile name, usually bottom left corner, 00:02:02.260 --> 00:02:05.079 then navigate to Settings. From there, head over 00:02:05.079 --> 00:02:07.340 to Data Controls and you'll see Shared Links. 00:02:07.739 --> 00:02:10.259 Just review that list carefully. For each link 00:02:10.259 --> 00:02:12.719 you find, click the little three dot icon, the 00:02:12.719 --> 00:02:14.759 vertical one, and select Delete Shared Link. 00:02:15.300 --> 00:02:19.219 Doing that makes the old URL return a 404 not 00:02:19.219 --> 00:02:21.530 found error. So it breaks the link. That's the 00:02:21.530 --> 00:02:24.729 start. OK. But a 404, while good, isn't always 00:02:24.729 --> 00:02:26.990 the complete end of the story, is it? Like the 00:02:26.990 --> 00:02:29.310 title or maybe a small snippet of that conversation 00:02:29.310 --> 00:02:31.310 might still hang around in search results for 00:02:31.310 --> 00:02:34.199 a bit, like an echo. Precisely. That echo is 00:02:34.199 --> 00:02:36.819 the problem. And that's where the second crucial 00:02:36.819 --> 00:02:39.479 step comes in. You need to actively speed up 00:02:39.479 --> 00:02:42.060 its complete removal from Google search index 00:02:42.060 --> 00:02:44.939 using their own tool. So after you've deleted 00:02:44.939 --> 00:02:48.120 the link inside ChatGPT, you go straight to Google's 00:02:48.120 --> 00:02:50.659 remove outdated content tool. Once you're there, 00:02:50.680 --> 00:02:53.360 you click new request, paste in that ChatGPT 00:02:53.360 --> 00:02:55.580 URL you just deleted, and just follow their instructions. 00:02:55.879 --> 00:02:58.259 They're pretty straightforward. Google usually 00:02:58.259 --> 00:03:01.520 processes these requests in maybe a few hours. 00:03:01.610 --> 00:03:03.669 Sometimes it can take up to two or three days 00:03:03.669 --> 00:03:05.550 for it to fully vanish from the search results, 00:03:05.750 --> 00:03:08.330 but it gets it done. So just to be crystal clear, 00:03:08.490 --> 00:03:10.569 then, if we're talking about the single most 00:03:10.569 --> 00:03:13.710 crucial non -negotiable action for a really complete 00:03:13.710 --> 00:03:16.590 cleanup, what is that? It's definitely a two 00:03:16.590 --> 00:03:18.990 -part dance. You absolutely have to delete the 00:03:18.990 --> 00:03:21.409 link inside ChatGPT first. But critically, you 00:03:21.409 --> 00:03:23.610 must follow that up by using Google's tool to 00:03:23.610 --> 00:03:25.550 tell them to remove the outdated content from 00:03:25.550 --> 00:03:27.689 their search results. You need both steps. One 00:03:27.689 --> 00:03:29.509 without the other isn't really a full cleanup. 00:03:29.629 --> 00:03:33.129 Right. Got it. Okay, so this chat GPT thing was 00:03:33.129 --> 00:03:35.250 no doubt a big deal, a really stark reminder 00:03:35.250 --> 00:03:37.909 about how fragile digital privacy can be sometimes. 00:03:39.330 --> 00:03:41.189 But connecting this to the bigger picture, it 00:03:41.189 --> 00:03:43.150 does feel like maybe just the tip of the iceberg, 00:03:43.310 --> 00:03:45.520 doesn't it? The risk isn't just about ChatGPT. 00:03:45.699 --> 00:03:48.400 Oh, absolutely. While ChatGPT got all the headlines, 00:03:48.560 --> 00:03:50.740 you know, other AI models like Google Gemini 00:03:50.740 --> 00:03:53.500 or Claude, they have their own ways of sharing 00:03:53.500 --> 00:03:55.580 their own data policies. You really need to look 00:03:55.580 --> 00:03:58.439 into those, too. But beyond just sharing, there 00:03:58.439 --> 00:04:00.979 are other risks, maybe more subtle ones, lurking 00:04:00.979 --> 00:04:03.800 around, like, for instance, how AI models use 00:04:03.800 --> 00:04:06.800 your data for training. Many models, sort of 00:04:06.800 --> 00:04:09.120 by default, might use your conversations to train 00:04:09.120 --> 00:04:12.180 future versions. Basically, the AI learns from 00:04:12.180 --> 00:04:14.669 your input to get better. unless you specifically 00:04:14.669 --> 00:04:17.550 go into settings and turn that off. Then you've 00:04:17.550 --> 00:04:19.649 got third -party extensions, these little browser 00:04:19.649 --> 00:04:22.269 add -ons that connect with AI. They can sometimes 00:04:22.269 --> 00:04:24.250 harvest your data without you really knowing 00:04:24.250 --> 00:04:26.629 exactly what they're taking. And finally, there's 00:04:26.629 --> 00:04:29.750 this idea of supply chain attacks. AI tools often 00:04:29.750 --> 00:04:31.769 rely on lots of other software libraries and 00:04:31.769 --> 00:04:34.610 services. A vulnerability anywhere in that chain, 00:04:34.790 --> 00:04:37.209 like one weak link, could potentially expose 00:04:37.209 --> 00:04:39.620 your data that flows through it. So it sounds 00:04:39.620 --> 00:04:41.939 like it's not just about how we choose to share 00:04:41.939 --> 00:04:44.319 information, but we need to be fundamentally 00:04:44.319 --> 00:04:47.680 aware of where our data is going and maybe who 00:04:47.680 --> 00:04:50.680 else might be seeing it or using it. Beyond the 00:04:50.680 --> 00:04:52.980 direct sharing risk, in your view, what's maybe 00:04:52.980 --> 00:04:54.980 the most significant hidden risk that people 00:04:54.980 --> 00:04:57.720 tend to overlook? That's a great question. I'd 00:04:57.720 --> 00:05:00.300 argue it's that AI training aspect we just touched 00:05:00.300 --> 00:05:02.819 on, the fact that models might be learning from 00:05:02.819 --> 00:05:05.019 your private conversations without you actively 00:05:05.019 --> 00:05:07.360 consenting or even realizing it's the default 00:05:07.360 --> 00:05:10.019 setting. Many platforms just use your inputs 00:05:10.019 --> 00:05:12.160 to improve themselves unless you opt out. That's 00:05:12.160 --> 00:05:15.259 a huge, often unseen data exposure right there. 00:05:15.680 --> 00:05:17.439 OK, this definitely raises a really important 00:05:17.439 --> 00:05:20.230 question then. Given all these risks, how do 00:05:20.230 --> 00:05:22.110 we fundamentally shift our approach? How do we 00:05:22.110 --> 00:05:25.310 actually build safer, more robust AI work habits? 00:05:25.470 --> 00:05:27.689 The source material gives us three sort of golden 00:05:27.689 --> 00:05:29.970 rules. Yeah, and these rules are really about 00:05:29.970 --> 00:05:32.050 changing your mindset. The first one is maybe 00:05:32.050 --> 00:05:34.930 the most profound. Treat every single input as 00:05:34.930 --> 00:05:37.550 if it's a potential public record. Before you 00:05:37.550 --> 00:05:39.649 type anything into an AI, just pause for a second, 00:05:39.670 --> 00:05:42.970 ask yourself, would I be okay with this information 00:05:42.970 --> 00:05:45.790 appearing on the front page of a newspaper tomorrow? 00:05:47.319 --> 00:05:50.220 If the answer's no, just don't enter it. It's 00:05:50.220 --> 00:05:53.300 a really simple but incredibly powerful mental 00:05:53.300 --> 00:05:56.120 check. That is a powerful filter. I have to admit, 00:05:56.199 --> 00:06:00.180 I still wrestle with prompt drift myself sometimes. 00:06:00.560 --> 00:06:02.360 You know, you get into a long chat and you start 00:06:02.360 --> 00:06:03.839 getting maybe a little too comfortable, a little 00:06:03.839 --> 00:06:06.560 too detailed. What's the next golden rule for 00:06:06.560 --> 00:06:09.319 us? The second rule is all about practical vigilance. 00:06:09.620 --> 00:06:12.000 Regularly audit your privacy settings, like really 00:06:12.000 --> 00:06:14.399 regularly. AI platforms update all the time and 00:06:14.399 --> 00:06:16.360 sometimes those updates quietly change default 00:06:16.360 --> 00:06:18.920 settings, so. Make it a habit, maybe monthly. 00:06:19.339 --> 00:06:21.120 Just pop into the settings sections, look for 00:06:21.120 --> 00:06:23.259 things like data controls, privacy, security, 00:06:23.319 --> 00:06:25.579 and your AI tools. Make sure everything still 00:06:25.579 --> 00:06:28.040 aligns with how you want your data handled. Things 00:06:28.040 --> 00:06:29.800 change fast, so your checks need to keep up. 00:06:30.139 --> 00:06:32.040 And the third rule, it sounds like it shifts 00:06:32.040 --> 00:06:33.980 the responsibility. It makes it broader than 00:06:33.980 --> 00:06:36.639 just say, the IT department's job. It's everyone's 00:06:36.639 --> 00:06:38.930 job now. That's exactly right. Digital hygiene, 00:06:39.129 --> 00:06:41.149 especially with AI tools, is now just a core 00:06:41.149 --> 00:06:43.649 part of modern work culture. It has to be. Whether 00:06:43.649 --> 00:06:45.930 you're a leader, a creator, an employee, whatever 00:06:45.930 --> 00:06:49.129 your role, actively protecting data, your own, 00:06:49.350 --> 00:06:52.569 your teams, your customers. It's just an essential 00:06:52.569 --> 00:06:55.370 skill now, non -negotiable. It's really everyone's 00:06:55.370 --> 00:06:57.889 shared responsibility to keep that digital environment 00:06:57.889 --> 00:07:01.290 secure. OK, so if we boil these rules down to 00:07:01.290 --> 00:07:04.269 a core message, for our daily habits, what's 00:07:04.269 --> 00:07:06.410 the essence we should carry forward? What's the 00:07:06.410 --> 00:07:08.250 takeaway? I think the essence is really a three 00:07:08.250 --> 00:07:10.370 -part commitment. First, think before you type 00:07:10.370 --> 00:07:13.329 anything sensitive. Second, regularly check your 00:07:13.329 --> 00:07:15.750 privacy settings. And third, always remember 00:07:15.750 --> 00:07:18.329 that data security is a team effort. Those are 00:07:18.329 --> 00:07:20.649 kind of the pillars for safe AI interaction, 00:07:20.990 --> 00:07:23.329 personal responsibility plus collective awareness. 00:07:23.829 --> 00:07:25.910 OK, so if sharing those live links directly from 00:07:25.910 --> 00:07:28.889 ChadGPD or similar platforms carries these inherent 00:07:28.889 --> 00:07:31.660 risks we've talked about. What's a safer, more 00:07:31.660 --> 00:07:34.540 reliable workflow we should actually adopt? Especially 00:07:34.540 --> 00:07:36.959 for sharing AI -generated ideas or information 00:07:36.959 --> 00:07:39.560 with colleagues. Yeah, great question. The main 00:07:39.560 --> 00:07:42.019 goal here is to completely break that direct 00:07:42.019 --> 00:07:45.370 link back to the original AI chat. The new, safer 00:07:45.370 --> 00:07:47.610 process is actually pretty simple though. Okay, 00:07:47.829 --> 00:07:50.449 it does add a few extra minutes. First step, 00:07:50.670 --> 00:07:52.470 obviously, generate your idea or content with 00:07:52.470 --> 00:07:55.689 the AI. Then, this is crucial, copy that content 00:07:55.689 --> 00:07:58.610 out of the AI and into a secure, controlled document. 00:07:58.670 --> 00:08:00.769 Something like Google Docs, Notion, maybe a standard 00:08:00.769 --> 00:08:03.759 Word doc. Then, and you must do this diligently, 00:08:04.040 --> 00:08:05.980 rigorously, redact and remove all identifying 00:08:05.980 --> 00:08:07.819 information. I mean, things like client names, 00:08:08.220 --> 00:08:09.860 specific financial numbers, internal project 00:08:09.860 --> 00:08:12.540 codes. Replace them all with generic placeholders 00:08:12.540 --> 00:08:15.439 like client name or project X. Only after that 00:08:15.439 --> 00:08:17.660 cleansing process do you share that secure document. 00:08:18.019 --> 00:08:21.279 Never. ever the original AI link itself. And 00:08:21.279 --> 00:08:23.339 hey, for stuff that's extremely sensitive, you 00:08:23.339 --> 00:08:25.459 can even go the route of taking screenshots and 00:08:25.459 --> 00:08:27.879 manually blacking out the critical bits. It's 00:08:27.879 --> 00:08:31.019 definitely less convenient, sure, but it is undeniably 00:08:31.019 --> 00:08:35.559 safer. It completely severs the tie. So the absolute 00:08:35.559 --> 00:08:37.919 safest approach for sharing AI insights really 00:08:37.919 --> 00:08:40.000 boils down to what? What's the core principle? 00:08:40.200 --> 00:08:42.799 It's adding that layer of rigorous manual review 00:08:42.799 --> 00:08:45.299 to the content first, and then sharing it only 00:08:45.299 --> 00:08:47.460 through secure, controlled channels you trust, 00:08:47.779 --> 00:08:50.620 never directly linking back to the raw AI conversation 00:08:50.620 --> 00:08:54.899 itself. Copy, cleanse, share securely. OK. We've 00:08:54.899 --> 00:08:56.840 talked about reacting to incidents after they 00:08:56.840 --> 00:08:59.460 occur, about building better habits moving forward. 00:09:00.000 --> 00:09:02.860 But proactive defense. Yeah. That feels like 00:09:02.860 --> 00:09:04.740 the real key to long -term security, doesn't 00:09:04.740 --> 00:09:07.820 it? What if we could integrate some kind of safety 00:09:07.820 --> 00:09:10.559 co -pilot directly into our AI workflow, something 00:09:10.559 --> 00:09:12.679 that reviews our inputs before we even hit send? 00:09:13.000 --> 00:09:15.000 Okay, now this is where it gets really interesting 00:09:15.000 --> 00:09:17.620 and honestly kind of mind -blowing. Our source 00:09:17.620 --> 00:09:20.480 material shares this incredibly detailed prompt 00:09:20.480 --> 00:09:23.179 template. And it's not just a simple command, 00:09:23.299 --> 00:09:26.100 you know, it's basically a full -blown risk management 00:09:26.100 --> 00:09:28.539 framework baked into a prompt. Think of it like 00:09:28.539 --> 00:09:30.879 a rapid safety audit you run before you'd have 00:09:30.879 --> 00:09:33.509 any sensitive interaction with the AI. The brain 00:09:33.509 --> 00:09:36.210 behind this prompt. It's built on three really 00:09:36.210 --> 00:09:39.730 solid pillars. First, the NIST AI RMF. That's 00:09:39.730 --> 00:09:42.009 the U .S. National Institute of Standards and 00:09:42.009 --> 00:09:44.850 Technology's AI Risk Management Framework. Basically, 00:09:45.009 --> 00:09:47.370 a structured government process for governing, 00:09:47.850 --> 00:09:50.730 mapping, measuring, and managing AI risks. The 00:09:50.730 --> 00:09:53.309 prompt actually simulates this process. Second, 00:09:53.590 --> 00:09:56.950 the OWASP LLM Top 10. That's the open worldwide 00:09:56.950 --> 00:09:59.470 application security projects list of the biggest 00:09:59.470 --> 00:10:01.710 security vulnerabilities for large language models. 00:10:02.250 --> 00:10:04.730 common attacks like prompt injection or leaking 00:10:04.730 --> 00:10:07.289 sensitive info, the prompt actively looks for 00:10:07.289 --> 00:10:10.679 these. And third, GDPR principles. You know, 00:10:10.860 --> 00:10:13.200 Europe's big data protection regulation, which 00:10:13.200 --> 00:10:15.840 really emphasizes data minimization, basically 00:10:15.840 --> 00:10:18.299 only collecting and keeping data that's absolutely 00:10:18.299 --> 00:10:21.100 necessary. The prompt pushes you towards that, 00:10:21.519 --> 00:10:24.460 recommending removing extra info. So by combining 00:10:24.460 --> 00:10:26.600 these three things, the prompt doesn't just check 00:10:26.600 --> 00:10:29.539 your text, it forces the AI to almost think like 00:10:29.539 --> 00:10:32.419 a security expert, a data privacy lawyer, and 00:10:32.419 --> 00:10:35.740 a risk manager all at once. Whoa. I mean, imagine 00:10:35.740 --> 00:10:37.779 scaling that kind of proactive defense across 00:10:37.779 --> 00:10:40.679 a whole organization. security right into the 00:10:40.679 --> 00:10:43.120 workflow from the start. That's genuinely powerful 00:10:43.120 --> 00:10:45.320 stuff for data governance. That sounds incredibly 00:10:45.320 --> 00:10:47.559 sophisticated in its design. Let's just briefly 00:10:47.559 --> 00:10:49.399 look in the prompt template structure. How does 00:10:49.399 --> 00:10:51.399 it actually function? Yeah, the structure is 00:10:51.399 --> 00:10:53.700 super clear, which is great. It starts with a 00:10:53.700 --> 00:10:57.759 system instruction. This tells the AI what its 00:10:57.759 --> 00:10:59.820 role is. Essentially, it becomes your personal 00:10:59.820 --> 00:11:01.919 safety and privacy reviewer for this interaction. 00:11:02.620 --> 00:11:05.440 Then you have the in PUSC section. This is where 00:11:05.440 --> 00:11:07.779 you fill in the key details for context. Your 00:11:07.779 --> 00:11:10.120 country, the purpose of the AI interaction, what 00:11:10.120 --> 00:11:12.639 kinds of data are involved, like PII, financial 00:11:12.639 --> 00:11:14.919 data, that sort of thing, how you plan to share 00:11:14.919 --> 00:11:17.179 the output, and importantly, your acceptable 00:11:17.179 --> 00:11:19.840 risk tolerance level. And then finally, you have 00:11:19.840 --> 00:11:22.440 the task itself. This tells the AI to execute 00:11:22.440 --> 00:11:25.080 five specific crucial steps based on your inputs. 00:11:25.279 --> 00:11:27.360 And those five steps, that's where the real analysis, 00:11:27.480 --> 00:11:29.399 the real safety check happens, right? Exactly. 00:11:29.559 --> 00:11:32.620 Each step does a specific job. First is pre -check 00:11:32.620 --> 00:11:35.440 map. This step scans your input and identifies 00:11:35.440 --> 00:11:38.519 any sensitive bits, PII, secrets, anything that 00:11:38.519 --> 00:11:40.740 might violate data minimization principles like 00:11:40.740 --> 00:11:43.980 under GDPR, and it even flags risks from that 00:11:43.980 --> 00:11:46.759 OWASP list like potential prompt injection vulnerabilities. 00:11:47.399 --> 00:11:49.759 Second step is measure. Here it actually scores 00:11:49.759 --> 00:11:51.820 each risk it found based on severity and likelihood. 00:11:52.220 --> 00:11:54.240 It presents this in a little table super easy 00:11:54.240 --> 00:11:57.200 to grasp like accidental PIR exposure, high risk 00:11:57.200 --> 00:12:00.840 or prompt injection, medium risk. Third is manage. 00:12:01.100 --> 00:12:03.580 This is the mitigation part. It gives you concrete 00:12:03.580 --> 00:12:06.000 suggestions for redactions and rewrites. It might 00:12:06.000 --> 00:12:08.379 replace names with placeholders like employee 00:12:08.379 --> 00:12:11.659 name, rewrite sentences to avoid revealing confidential 00:12:11.659 --> 00:12:14.440 figures, maybe reduce the granularity of data, 00:12:14.779 --> 00:12:17.759 or add constraints like do not ask for home addresses. 00:12:18.360 --> 00:12:21.279 Fourth is safety controls. Based on your context 00:12:21.279 --> 00:12:23.620 and region, it adds five specific guardrails, 00:12:23.740 --> 00:12:26.559 things like ensure no external links are executed 00:12:26.559 --> 00:12:29.980 by the AI or limit data retention and logging 00:12:29.980 --> 00:12:32.779 for this specific chat session. tailored advice. 00:12:33.360 --> 00:12:35.559 And finally, step five is the final gate. This 00:12:35.559 --> 00:12:37.620 is the bottom line. It gives you one of three 00:12:37.620 --> 00:12:40.740 outputs, either a clean pumped output, safe version, 00:12:40.740 --> 00:12:42.960 ready to use, or a concise checklist of things 00:12:42.960 --> 00:12:45.539 you still need to do manually. Or if it decides 00:12:45.539 --> 00:12:47.679 the risk is still too high, even after mitigations, 00:12:47.759 --> 00:12:49.960 it just says do not use and briefly explains 00:12:49.960 --> 00:12:52.240 why. OK, let's try to make this really tangible. 00:12:52.340 --> 00:12:55.240 Could you walk us through how, say, an HR professional 00:12:55.240 --> 00:12:57.360 might use this in their day to day work? Perfect 00:12:57.360 --> 00:13:01.860 example. OK, imagine an HR or maybe a legal professional, 00:13:01.860 --> 00:13:03.659 let's say, based in Vietnam, they need to use 00:13:03.659 --> 00:13:06.220 an AI to help summarize an employee complaint 00:13:06.220 --> 00:13:08.379 before briefing internal leadership. So their 00:13:08.379 --> 00:13:11.059 inputs would be country, Vietnam, context, employee 00:13:11.059 --> 00:13:13.179 complaint summary for leadership briefing, data 00:13:13.179 --> 00:13:16.139 types, PII, sensitive allegations, sharing, internal 00:13:16.139 --> 00:13:18.820 leadership only, risk tolerance, very low because 00:13:18.820 --> 00:13:21.419 it's sensitive HR data. Okay, so they run the 00:13:21.419 --> 00:13:23.960 prompt. The output would likely include a risk 00:13:23.960 --> 00:13:26.919 table highlighting the high risk of PII exposure 00:13:26.919 --> 00:13:29.340 and maybe potential legal implications if not 00:13:29.340 --> 00:13:32.539 handled correctly. Then the manage step would 00:13:32.539 --> 00:13:35.220 generate a carefully anonymized summary. It would 00:13:35.220 --> 00:13:37.580 replace specific names, like maybe Ning Yen Ven 00:13:37.580 --> 00:13:40.860 Eh, with a placeholder like complainant. Specific 00:13:40.860 --> 00:13:43.460 dates or times might become vaguer, like early 00:13:43.460 --> 00:13:46.360 August. And the accompanying checklist, the final 00:13:46.360 --> 00:13:48.539 gate part, would probably include crucial points 00:13:48.539 --> 00:13:51.240 like Share this summary only through encrypted 00:13:51.240 --> 00:13:53.659 email channels, store the original complaint 00:13:53.659 --> 00:13:56.059 document on a secure, access -controlled server, 00:13:56.320 --> 00:13:59.179 and critically, ensure the AI chat history for 00:13:59.179 --> 00:14:01.860 this specific session is disabled and then deleted 00:14:01.860 --> 00:14:04.899 immediately after use. You see, it's not just 00:14:04.899 --> 00:14:07.159 cleaning the text, it's building in these essential 00:14:07.159 --> 00:14:09.840 process safeguards around the interaction. Wow. 00:14:10.039 --> 00:14:11.860 So it really is like having a digital lawyer 00:14:11.860 --> 00:14:13.779 and a security expert built into your process, 00:14:13.860 --> 00:14:16.639 reviewing your AI interactions before any sensitive 00:14:16.639 --> 00:14:18.750 data even leaves your control. Yeah. What would 00:14:18.750 --> 00:14:21.129 you say is the ultimate most profound benefit 00:14:21.129 --> 00:14:22.909 of using a prompt like this for our listeners? 00:14:23.289 --> 00:14:26.230 The ultimate benefit is truly gaining an AI co 00:14:26.230 --> 00:14:29.710 -pilot dedicated to proactive security. It embeds 00:14:29.710 --> 00:14:31.929 a really sophisticated risk assessment right 00:14:31.929 --> 00:14:34.409 into your daily workflow. It helps transform 00:14:34.409 --> 00:14:37.070 AI from what could be a potential liability into 00:14:37.070 --> 00:14:39.629 a securely managed powerful asset. That's the 00:14:39.629 --> 00:14:42.769 core value, mid -roll sponsor read. So when we 00:14:42.769 --> 00:14:44.570 step back and look at all this, what does it 00:14:44.570 --> 00:14:47.429 really mean for us, this chat GPT incident? It 00:14:47.429 --> 00:14:49.490 feels like it was a pivotal moment, maybe even 00:14:49.490 --> 00:14:51.710 an expensive lesson for this digital age we're 00:14:51.710 --> 00:14:54.419 in. An expensive lesson, maybe, but a truly valuable 00:14:54.419 --> 00:14:56.320 one, I think. The big idea, the main takeaway 00:14:56.320 --> 00:14:59.460 we want you to hold onto is this. AI is an unbelievably 00:14:59.460 --> 00:15:02.120 powerful tool. It's a genuine amplifier of human 00:15:02.120 --> 00:15:04.460 capability, no doubt about it. But with that 00:15:04.460 --> 00:15:07.320 immense power comes an equally immense responsibility. 00:15:07.419 --> 00:15:09.980 It really is on us, the users, to be vigilant, 00:15:10.279 --> 00:15:12.700 to truly understand the dynamic digital landscape 00:15:12.700 --> 00:15:14.879 we're operating in, and to take active conscious 00:15:14.879 --> 00:15:18.600 ownership of our data. Yes, exactly. By proactively 00:15:18.600 --> 00:15:21.179 cleaning up our digital footprints, like we discussed, 00:15:21.360 --> 00:15:24.759 by adopting these much safer workflows, and by 00:15:24.759 --> 00:15:27.580 consciously building a truly vigilant privacy 00:15:27.580 --> 00:15:31.360 -first mindset. We really can harness AI's incredible 00:15:31.360 --> 00:15:34.299 potential without constantly worrying about sacrificing 00:15:34.299 --> 00:15:37.120 our privacy and security. It's all about intelligent 00:15:37.120 --> 00:15:40.220 engagement, isn't it? Informed, responsible engagement. 00:15:40.360 --> 00:15:43.000 Couldn't agree more. It really means embracing 00:15:43.000 --> 00:15:45.679 the innovation that AI offers, wholeheartedly, 00:15:45.960 --> 00:15:48.940 but doing it while maintaining unwavering conscious 00:15:48.940 --> 00:15:51.899 control over our data. It's about finding that 00:15:51.899 --> 00:15:54.379 balance. This deep dive has hopefully shed some 00:15:54.379 --> 00:15:56.139 light on both the immediate actions you might 00:15:56.139 --> 00:15:58.299 need to take and the kind of strategic mindset 00:15:58.299 --> 00:16:01.250 required for a more secure AI future. Now as 00:16:01.250 --> 00:16:02.730 we wrap up, we want to leave you with a thought 00:16:02.730 --> 00:16:05.509 to maybe mull over. Yeah, what other hidden risks 00:16:05.509 --> 00:16:07.909 might still be lurking just beneath the surface 00:16:07.909 --> 00:16:10.429 as AI continues to evolve at this absolutely 00:16:10.429 --> 00:16:12.990 breakneck pace? And maybe more importantly, think 00:16:12.990 --> 00:16:15.009 about how you personally will integrate this 00:16:15.009 --> 00:16:17.549 newfound vigilance, this awareness, into your 00:16:17.549 --> 00:16:20.429 own daily digital interactions from now on. Definitely 00:16:20.429 --> 00:16:22.710 something to consider. Food for thought indeed. 00:16:22.950 --> 00:16:24.809 Thank you so much for diving deep with us today. 00:16:25.009 --> 00:16:28.429 Until next time, stay curious and above all, 00:16:28.649 --> 00:16:31.289 stay safe out there. OTR Music.