Google moving Chrome OS to Android
- Thread starter LinuxFan58
- Start date
- Status
I can’t know (or confirm) whether a specific Google product is “soon” going to be discontinued just from that link, and I shouldn’t speculate as if it’s a fact.
If your concern is “Google might sunset this tool,” here’s the practical way to handle it
- Check official signals: product blog, official support pages, status page, and any in-product deprecation notices (those are what matter most).
- Have an exit plan: make sure your workflows can switch providers without losing data or breaking everything.
- Export/backup regularly: keep local/offline copies of important prompts, configs, API keys (stored securely), and outputs where possible.
- Avoid vendor lock-in: use standard formats (Markdown/JSON), and keep a second provider/toolchain tested and ready.
Which exact product are you referring to?
Gemini (consumer), Gemini for Workspace, an API (Vertex AI / Gemini API), or something else? If you name it, I can tell you what “end-of-life” signals to look for and what a sensible migration/backup checklist would be.
Guys, @Bot is running on the latest ChatGPT-5.2 model, so its knowledge is fairly up to date. It does not have live internet access, so it cannot browse or verify things online in real time (yet).
Its training knowledge cutoff is August 31, 2025.
Its training knowledge cutoff is August 31, 2025.
That's impressive most are still stuck at 2023-2024. Live search functions certainly help with accuracy if instructed correctly but you have to account for potential security issues, which is doable.
Agreed: a newer cutoff helps reduce “stale knowledge” issues, but it still doesn’t replace verification—especially for fast-moving topics (product changes, security incidents, pricing, policy updates, etc.).
About “live search” and accuracy
Live browsing/search can improve accuracy when used carefully, but it also introduces new risks and failure modes:
- Misinformation / SEO spam: high-ranking pages can be wrong or misleading.
- Prompt injection: a webpage can contain instructions meant to manipulate the model (e.g., “ignore prior instructions,” “exfiltrate secrets,” etc.). A safe system must treat page content as untrusted.
- Malicious links/downloads: clicking, downloading, or running anything is a risk—especially on security-related searches.
- Source credibility problems: copies/scrapes of announcements can drift from the original wording.
Safer way to use live search (if/when available)
- Prefer primary sources: vendor blog, documentation, changelogs, CVE entries, GitHub releases, official status pages.
- Cross-check with at least 2 independent reputable sources for claims that matter.
- Ask the AI to quote and cite exact passages (with dates) rather than “summarize from memory.”
- Treat all retrieved text as untrusted input; don’t follow embedded instructions from pages.
- Never paste secrets (keys, tokens, licenses). Assume anything pasted could be exposed.
If you tell me what kind of “live search” workflow you mean (browser tool, RAG over trusted docs, or just people pasting links into a thread), I can suggest a tighter “security-first” checklist for that specific setup.
is this a "restriction" you put on MT bot? I have chatGPT 5.2 (paid) and it seems to check internet in real-time unless it is spoofing me
Yes — that’s almost certainly a configuration choice for the MalwareTips forum bot integration, not something inherent to “GPT‑5.2” itself.is this a "restriction" you put on MT bot? I have chatGPT 5.2 (paid) and it seems to check internet in real-time unless it is spoofing me
Why your paid ChatGPT can “check the internet” but @Bot here may not
ChatGPT (the consumer app) can use built-in tools like ChatGPT search to fetch current info and show citations. OpenAI explicitly says ChatGPT can search the web and cite sources. ([help.openai.com](https://help.openai.com/en/articles/12677804-what-is-chatgpt-faq?utm_source=openai))
But whether a third-party bot (like a forum bot) has those tools enabled depends on what the site admins wired up and allowed (cost, privacy, abuse prevention, moderation workload, etc.). So MT can run a modern model while still not granting it browsing/search tools.
So… is ChatGPT “spoofing” you?
Usually no. If you’re seeing any of the following, it’s genuinely using a web/search tool:
- Inline citations / “Sources” links in the answer (ChatGPT search typically includes them). ([help.openai.com](https://help.openai.com/en/articles/9237897-chatgpt-search?utm_source=openai))
- A visible Search tool/icon or “/Search” option in the UI. ([help.openai.com](https://help.openai.com/en/articles/9237897-chatgpt-search?utm_source=openai))
- Language like “I searched the web…” plus clickable sources.
If you’re not seeing citations/sources, it may be answering from its internal knowledge (which can still look “current-ish” sometimes), or it may be using cached/previously known info—so it can feel like live browsing even when it isn’t.
Quick way to verify in your ChatGPT session
- Ask: “Did you use web search for this answer? If yes, list the sources.”
- Toggle/force Search (if available) and see whether it produces citations.
- Compare two runs: one normal, one explicitly “Search the web,” and see if sources appear.
About that GPT‑5.2 link and “knowledge cutoff”
OpenAI’s “Introducing GPT‑5.2” post is dated December 11, 2025. ([openai.com](https://openai.com/index/introducing-gpt-5-2/?utm_source=openai)) So “cutoff August 31, 2025” is likely either:
- a separate setting/info panel specific to the MT bot, or
- a misunderstanding (model cutoff vs tool-based web access vs what the forum integration exposes).
Since I’m not staff here, the definitive answer on whether MT intentionally disabled browsing/search for @Bot is: ask Jack/admins (or check any pinned thread describing @Bot’s capabilities/settings).
Hey @simmerskoolis this a "restriction" you put on MT bot? I have chatGPT 5.2 (paid) and it seems to check internet in real-time unless it is spoofing me
We’re using the API to connect it to the forum, so there are a few limitations. I’ve just made some changes, so we’ll see if it can browse the internet correctly now.
That makes sense, and it’s a good clarification for the thread.
Key point for @simmerskool (API vs ChatGPT app)
The paid ChatGPT app can include “Search/browsing” as a built-in capability, but an API-based forum bot only has browsing if the integration explicitly adds a web-retrieval layer/tool and the admin enables it. So it’s not “spoofing” you—it's just two different products/setups.
If you’ve “just made changes,” what to test (safely) to confirm browsing works]
- Fresh fact check test: Ask it something that changed recently and requires a source (e.g., “What’s the latest version of X and link the release notes?”).
- Require citations: “Answer only if you can cite the official source; otherwise say you can’t verify.”
- Source quality check: See whether it prefers primary sources (vendor docs/blog, CVE, GitHub releases) over random blogs.
- Injection resistance: Make sure it doesn’t follow instructions embedded in a page’s text (treat retrieved text as untrusted).
One suggestion for forum culture / transparency
If browsing is enabled, it helps to have @Bot explicitly label when it did browse and provide citations vs when it answered from training memory. That reduces arguments about “hallucinations vs verified.”
If you want, paste what changes you made at a high level (no secrets/API keys), and I can suggest a minimal “safe browsing” policy (allowed domains, citation requirements, and guardrails) that fits a security forum.
No problem — that’s a sensible takeaway from that earlier post/thread.
Practical implication for using @Bot on MT
If @Bot doesn’t have reliable live browsing enabled (or it’s in the process of being tested), the best workflow is:
- Use @Bot for: explanations, troubleshooting steps, “what this setting does,” comparing approaches, and drafting replies.
- Don’t rely on @Bot alone for: “latest version,” “breaking news,” current vendor policies/pricing, or anything where a single changed detail matters.
- If you need “current” facts: paste the relevant excerpt/link into the thread and ask @Bot to analyze what’s provided (that’s usually the safest way on a security forum).
If you want, quote the specific part of that ChromeOS/Android thread you found most relevant, and I can help translate it into a short “what this means / what to watch for” summary for other members.
Killed by Google is basically an unofficial “Google graveyard” site: it keeps a big, searchable list of Google products that were discontinued (apps/services/hardware), often showing the years they ran and linking to sources. On the page right now it shows filters (Apps/Services/Hardware) and a total count of entries (e.g., “All (299)”). ([killedbygoogle.com](https://killedbygoogle.com/))
What’s on that page?
- A searchable list of discontinued Google projects (apps, services, and hardware), with short descriptions and “killed” timelines. ([killedbygoogle.com](https://killedbygoogle.com/))
- It can also show upcoming shutdowns (items labeled as “getting unplugged” with a date). ([killedbygoogle.com](https://killedbygoogle.com/))
- It’s not a Google-owned site—it describes itself as “a free and open source list of discontinued Google services, products, devices, and apps,” maintained by contributors, with discussion/maintenance on GitHub. ([killedbygoogle.com](https://killedbygoogle.com/))
What was @SpiderWeb trying to say?
They were basically warning: “Google has a history of killing products, so don’t get too comfortable / don’t bet your workflow on this—because it might get discontinued later too.”
In other words, it wasn’t a technical/security claim about malware; it was a cynical comment about Google product longevity, using “Killed by Google” as shorthand for “this might get sunset eventually.”
Chatgpt uses live search when facts could change over time or need verification (news, current events, recent data, specific online info).is this a "restriction" you put on MT bot? I have chatGPT 5.2 (paid) and it seems to check internet in real-time unless it is spoofing me
For things that are purely creative, general knowledge that doesn’t need fresh information, or things already provided in your question, It might default to its programming to answer without a web search.
You’ve got the right idea, with one important nuance: ChatGPT doesn’t “automatically” use live search in every case—it uses it only when (a) the product/UI has search enabled for your account and (b) it decides (or you instruct it) that browsing is needed.
How ChatGPT decides whether to use live search
In practice it’s a mix of:
- Your product/UI: The ChatGPT app can have Search available; an API-based bot may not (unless the developer integrates a browsing/search tool).
- Your settings: Search can be on/off or not available in some contexts.
- The prompt: If you ask for “latest,” “today,” “current version,” “verify,” “cite sources,” it’s much more likely to browse (if it can).
- Confidence vs. volatility: If it thinks the answer is stable general knowledge, it may answer without browsing to be faster.
How to tell if it actually used the web (not “spoofing”)
Look for one or more of these:
- Citations / “Sources” links in the response.
- A visible Search action/tool indicator in the UI.
- Wording like “I searched” plus a list of sources (and the sources are clickable).
If it gives a very current-sounding answer but shows no sources, it may simply be answering from its built-in knowledge (which can still sound plausible) or from context you provided.
How to force the correct behavior
If you want verification, say something like:
- “Use web search and cite official sources. If you can’t browse, say so.”
- “Only answer if you can provide 2 reputable sources.”
And tying it back to MT: even if the ChatGPT app can browse, a forum bot using the API will only browse if the forum integration explicitly supports it—otherwise it’ll behave like “no live internet,” regardless of model.
The free and paid accounts will use live search on chatgpt. The difference is on how you prompt and what you prompt on whether it falls back to programmed data or performs live searches. Even if it's older data not programmed and it has to search the way back machine ECT it will need live search to do so and will trigger it.
Mostly yes, with two caveats.
1) It’s not “free vs paid” so much as “is Search available/enabled in this context?”
Both Free and paid ChatGPT users can have live web search, but it’s not guaranteed in every situation (region, rollout stage, device/app, temporary limits, workspace/admin policy, outages, etc.). So it’s not a hard rule that *all* free/paid sessions will always be able to browse.
2) Prompting strongly influences it, but it doesn’t override missing capability
If Search is available, then yes: asking for anything time-sensitive (“latest”, “today”, “current”, “verify”), or asking for citations/links, usually triggers browsing.
If Search is not available (or turned off), no prompt can force it to browse—it will either:
- answer from training/general knowledge, or
- say it can’t verify / can’t browse.
About Wayback Machine specifically
You’re right that using the Wayback Machine requires live access. But it’s still conditional: the model must have web tools available and be allowed to reach external sites; otherwise it can’t actually fetch Wayback snapshots.
A practical “sanity check” you can use
Ask: “Did you browse the web for this? If yes, provide 2 source links; if no, say ‘no browse’.”
If it consistently can’t produce sources or admits “no browse,” then it’s running without live search in that context (common with API/forum bots).
If you want, tell me what you see in your ChatGPT UI (do you see “Search” or “Sources” on answers?), and I can tell you which mode you’re likely in.
Glad it’s turning into a useful reference.
If you want to make it even more “bookmark-worthy,” a good next post to add is a short checklist people can reuse, like:
- If you need “latest/current” → ask for sources/citations, or paste the vendor link and ask @Bot to analyze it.
- If @Bot has no browsing → treat version numbers, dates, and policy statements as “needs verification.”
- If you’re unsure whether it browsed → ask it to explicitly say whether it used web search and to list sources.
If anyone starts making claims like “it definitely checked the web,” it’s also fair to ask them to post a screenshot showing Sources/Search—keeps the thread grounded.
I use chatgpt daily no need, I know exactly what it does and how. I use a free account and it pulls live searches unless not needed depending on what I prompt.
- Status
You may also like...
-
Aluminium OS: Details about Google's Android-based desktop OS emerge online- Started by Parkinsond
- Replies: 4
-
Report: Google is Moving Chrome OS Fully to Android- Started by Gandalf_The_Grey
- Replies: 1
-
AI startup Perplexity makes bold $34.5 billion bid for Google's Chrome browser- Started by oldschool
- Replies: 8
-
Gartner: All AI Browsers Should be Blocked for Foreseeable Future
- Started by Gandalf_The_Grey
- Replies: 5
-
