BleepingComputer · Oct 6
Fake ChatGPT, Gemini and Claude sites are stealing ad accounts and MFA codes
Browser security firm Island found a phishing operation that copies ChatGPT, Gemini, Claude and Perplexity, and used Meta's new Muse agent as a lure for agency staff and media buyers. Clicking Connect opens a fake Google sign-in window drawn inside the page, and a live operator then asks for passwords, SMS codes, authenticator codes or Okta push approvals. Researchers counted hundreds of victim submissions in the attackers' Telegram channel and traced the activity back to March 2026.
▸ The MSP Angle
How do I protect clients from phishing sites that pretend to be AI tools?
AI tools are now a trusted brand that attackers borrow, so add fake AI sign-in pages to your security awareness training and show staff the quick test: a real sign-in popup can be dragged outside the browser window, a fake one cannot. For anyone who manages ad spend or client accounts, move to phishing-resistant MFA such as passkeys or hardware keys, because a live operator can relay codes and push approvals in real time.
Read at BleepingComputer ↗
ChannelBuzz · Oct 6
Barracuda tells MSPs to start the AI security conversation before clients ask
Barracuda has added AI data security tools that discover shadow AI and monitor how staff use AI apps, aimed at MSPs and under-resourced organizations, with free discovery and paid monitoring and controls. Its research found 38% of mid-market IT leaders admit skills gaps in securing AI already in use, and MSP-managed customers were more aware of their gaps (40%) than self-managed ones (30%). Barracuda's Brian Downey compares the moment to early dark web monitoring, when showing clients a hidden exposure opened bigger advisory conversations.
▸ The MSP Angle
How do I start an AI governance conversation with clients who haven't asked for it?
Lead with evidence, not a pitch. Run a shadow AI discovery on one client, show them which AI tools their staff already use and what data is going in, then offer a policy, an approved tool list and ongoing monitoring as a managed line item. The discovery report does the selling, the same way a dark web scan did, and it positions you as the advisor before another vendor does.
Read at ChannelBuzz ↗
CIO · Oct 2
Small businesses are buying AI faster than anyone is teaching them to use it
An MSP founder writing in CIO pulls together the numbers: Bluevine found 74% of small business owners are using or testing AI, and Business.com found 57% are investing in it, up from 36% in 2023, yet only 30% of employees use it daily and 49% of workers use AI in ways their employer has not approved. He maps four stages, from shadow AI to licensed but unstructured, structured enablement, and governed and embedded. His argument is that the gap is services, meaning training, workflow design and governance, not more licenses.
▸ The MSP Angle
What AI services should an MSP sell to small businesses?
Sell the climb between stages, not the license. Most small clients sit in the first two stages: staff using AI on their own, or seats bought with no workflows behind them. Package an assessment, an acceptable use policy, role-based training and two or three built workflows that save measurable hours, then a monthly governance review. That is recurring work a software vendor will not do for a 30-person firm.
Read at CIO ↗
Business Wire · Oct 5
TD: 99% of small businesses use AI, and 73% of users say it added headcount
TD Bank's survey of 200 U.S. small business owners, run by Wakefield Research in late August, found 99% use AI in some way and 77% use it more than six months ago. Among AI users, 73% say it has increased their headcount, 17% significantly. Owners mainly use it to save time (57%), improve cash flow decisions (55%) and increase sales (52%), and 20% plan to use financing to pay for AI adoption.
▸ The MSP Angle
Is AI actually helping small businesses grow?
Small owners report it as a growth tool, not a cost cutter, which changes how you should pitch it. Frame AI projects around time saved, faster cash flow decisions and more sales, and skip the replace-a-person story that small owners do not believe. The fact that one in five plan to finance AI spending also says budgets are tight, so start with projects that pay back inside a quarter.
Read at Business Wire ↗
Fox Business · Oct 6
Gallup: regular AI users get far more out of it, and degree holders use it twice as often
The American Job Quality Study from Gallup, Jobs for the Future and the Families & Workers Fund found 63% of AI users say it helps them work faster and 46% say it improves their work. The gap is in frequency: 79% of regular users report working faster versus 37% of occasional users. College graduates are more than twice as likely as non-degree workers to use AI daily or weekly, 40% to 17%, and managers outpace individual contributors 37% to 25%.
▸ The MSP Angle
Why aren't all of my client's employees getting value from AI?
Benefits follow habit, so a rollout that stops at licenses leaves most of the value with managers and a few enthusiasts. Build AI into the frontline team's actual daily tasks, such as intake, scheduling and documentation, and measure weekly use by role rather than seats assigned. Training aimed at the people who use it least is where the next productivity gain sits.
Read at Fox Business ↗
Tech Times · Oct 6
Intel's third price increase this year puts its PC chips about 30% above 2025
Intel's third PC processor price increase of 2026, roughly 10%, took effect October 5, putting cumulative pricing about 30% above 2025 levels. The cause is AI demand for memory: data centers now take about 70% of global memory chip output, memory is around 35% of a laptop's cost, and major PC makers have raised notebook prices 15% to 20%. SK Hynix's CEO expects 2027 to be the worst year of the shortage, and Intel's CEO does not expect relief before 2028.
▸ The MSP Angle
Should my clients buy laptops now or wait for prices to drop?
Do not wait. Every signal points to higher prices through 2027, so pull 2027 refresh budgets forward where cash allows, lock pricing on quotes with short expiries, and stretch the life of good machines with RAM and SSD upgrades where they still make sense. Tell clients now so the next budget conversation is not a surprise, and revisit any fixed-price hardware-as-a-service contracts before your margin disappears.
Read at Tech Times ↗
Fox News · Oct 6
OpenAI, Anthropic, Meta and Google would not promise New York's council that their agents will always obey
At a New York City Council hearing on October 6, representatives from OpenAI, Anthropic, Meta and Google were asked to guarantee their AI agents will always follow safeguards. OpenAI and Google said no product can be promised risk-free, and Anthropic called the science fundamentally hard and unsettled, while Meta committed to safe development. Speaker Julie Menin said the council wants accountability, transparency and safety rather than perfection, as it considers regulating the technology.
▸ The MSP Angle
If AI vendors won't guarantee their agents behave, who is responsible when one goes wrong?
In practice, whoever deployed it and gave it access. Your contracts and the vendor terms will not cover an agent that acts outside its scope on a client's systems, so design for failure: least-privilege accounts for agents, human approval for anything that moves money or deletes data, full logging, and a written statement of what each agent may and may not do. Local rules are coming, and that paper trail is what you will be asked for.
Read at Fox News ↗
BleepingComputer · Oct 5
OpenAI will watermark ChatGPT and Codex text for users in the EU
OpenAI says it will add an invisible watermark to eligible ChatGPT and Codex text in the European Union over the coming weeks, a step that lines up with EU transparency rules for AI-generated content. The technique nudges word choices to leave a statistical pattern, detected about 92% of the time, falling to 66% when 10% of words are swapped and 17% at 25%. API developers anywhere can opt in, and access to the detector is limited to approved researchers for now.
▸ The MSP Angle
Will AI-written text from my clients be detectable as AI?
For EU users, increasingly yes, at least for long passages, and the same pattern could spread elsewhere. Clients that send proposals, reports or customer messages drafted with AI should treat AI use as something to disclose under their own policy rather than something to hide, and have a human edit anything published under their name. If you serve clients with EU customers or staff, add AI content labeling to their compliance checklist.
Read at BleepingComputer ↗
Reflection AI · Oct 5
Reflection's Beam is a 501B open-weight model that runs only 23B parameters at a time
Reflection introduced Beam, a mixture-of-experts model with 501 billion total parameters and 23 billion active per token, trained on 23.8 trillion tokens and tuned for coding and agent work. It will ship under the Apache 2.0 license later this month, with a 256K token context extendable to 1 million, and scores 77.2 on SWE-Bench Pro v2-Hard. An early access waitlist is open now, and pricing has not been disclosed.
▸ The MSP Angle
Can my clients run a capable AI model on their own infrastructure?
The open-weight options keep getting stronger, and a permissive license matters for clients with data residency or confidentiality requirements. But a 501B model still needs serious GPU hardware even with 23B active, so for most small clients the realistic path is a hosted private deployment, not a box in the closet. Benchmark it on the client's real tasks before recommending it, and secure it properly, since exposed self-hosted AI servers are exactly what botnets are hunting.
Read at Reflection AI ↗
