Lamont Information Technology

Lamont Information Technology Security-First IT for South African Business. Family-owned since 1999.

From outsourced CISO services and Microsoft 365 hardening to AI governance and day-to-day support that just works — we keep your business protected, compliant and running.

This week, AI wrote the patch list, and the attack.Here's what actually mattered in AI for IT (14–20 July)Update on our ...
20/07/2026

This week, AI wrote the patch list, and the attack.
Here's what actually mattered in AI for IT (14–20 July)
Update on our last post: the gate finally opened. In our 6 July post we said OpenAI's GPT-5.6 was still locked to 20 vetted partners, with a public launch "expected before the end of July." It landed early: GPT-5.6 (Sol, Terra, Luna) is now rolling out publicly via ChatGPT and the API. The lockdown then launch cycle we've tracked since 30 June has run its full course.
The "cheaper AI" race just had a second DeepSeek moment. Two weeks ago we flagged that serious AI is getting dramatically cheaper, pointing at China's low-cost models. This week Moonshot AI released Kimi K3 (16 July), a 2.8-trillion-parameter model billed as the largest open-source model ever, benchmarking neck and neck with the top systems from OpenAI and Anthropic. It rattled AI and chip stocks, and the full weights go public on 27 July. Frontier grade AI is getting cheaper and more open fast.
AI found a record 570 Windows flaws in a single month. Microsoft's July Patch Tuesday fixed 570 vulnerabilities its largest update ever and credits AI-assisted scanning for the surge. Three are zero-days (two already being exploited), including privilege escalation bugs in SharePoint and Active Directory Federation Services, plus a 9.6-severity remote code ex*****on flaw in Copilot itself. If you run Microsoft 365, patching just became a bigger, faster-moving job.
Researchers: AI has gone from "assistant" to "operator." Check Point's AI Security Report 2026 (14 July) warns that AI now runs live attacks writing malware, powering phishing-as-a-service, and driving a roughly 5x rise in prompt-injection attempts. It also found the average business quietly runs ~10 AI apps a month, many unsanctioned, with risky, data-exposing prompts doubling to 4%. Shadow AI is now a boardroom risk:
Update on Microsoft's "multi-model" bet. Two weeks ago we noted Microsoft's $2.5-billion "Frontier" push toward a multi-model world. This week the strategy sharpened: Microsoft is reportedly coaching its salesforce to pitch its own in-house models as cheaper and more efficient than OpenAI, Google, and Anthropic "selling the full end-to-end system." Multi-model, sure, but every vendor still wants to be your default.
What it means for SMEs: The same AI that's getting cheaper and more capable is also finding, and exploiting, security holes faster than ever. The smart play hasn't changed: patch fast (that M365 update is bigger this month for a reason), get visibility on the AI tools your team is already using, keep your data governed and portable, and harden your Microsoft 365 tenant before you stack AI on top of it.
That's the work we do at LITSA. Let's talk.

Last week AI got locked down. This week, the gates cracked open.In our 30 June post, we reported that Washington had cla...
08/07/2026

Last week AI got locked down. This week, the gates cracked open.
In our 30 June post, we reported that Washington had clamped down on the most powerful AI models — OpenAI's GPT-5.6 limited to a handful of vetted partners, and Anthropic's frontier models restricted to a small group of approved organisations. One week on, the story has already moved.
Here's what changed
Update on last week: The US Commerce Department lifted its export curbs on Anthropic's Claude Fable 5 and Mythos 5 — less than three weeks after the models were forced offline. Fable 5 returned globally on 1 July; Mythos re-opened to approved organisations. The lesson from last week now cuts both ways: frontier AI access can be switched off — and back on — with almost no notice. If your business is building on a single model, that's a risk to plan for.
GPT-5.6 is still behind the gate. Access remains limited to roughly 20 partners, with a public launch now expected before the end of July. The "coming weeks" we flagged last week are still counting.
The race is getting cheaper. Anthropic launched Claude Sonnet 5 (30 June) — near top-tier performance at a fraction of the price — and China's low-cost GLM-5.2 is now matching leading models on coding tasks. Serious AI is becoming dramatically more affordable for everyday business use.
Microsoft is betting on a "multi-model" world. It launched a new $2.5-billion "Frontier" business (2 July) to help companies pick, adapt and deploy the right mix of AI — a clear signal that no single vendor wins, and that AI strategy is now a boardroom decision.
What it means for SMEs: The smart play isn't chasing the newest model — it's staying flexible. Don't lock your business to one AI vendor, keep your data governed and portable, and get the security fundamentals right first: harden your Microsoft 365 tenant, control who (and what) can access your systems, and treat AI as a governed part of your stack.
That's the work we do at LITSA. Let's talk.

________________________________________

US Governments are now in the loop. OpenAI released its new GPT-5.6 models — but only to a small group of vetted partner...
30/06/2026

US Governments are now in the loop. OpenAI released its new GPT-5.6 models — but only to a small group of vetted partners, at the request of the US government. Anthropic's frontier "cyber" model, Claude Mythos 5, was cleared by the US Commerce Department for release to roughly 100 approved companies and agencies. Access to the most capable AI is becoming a privilege, not a default.
Why the caution? These models are now capable of finding and exploiting software vulnerabilities faster than any human analyst — which cuts both ways for defenders and attackers.
AI is reshaping the IT workforce. Oracle confirmed it has cut around 21,000 roles (~13%) over the past year, citing AI, and analysts say AI was the most-cited reason for tech-sector layoffs in May.
Following up on our last post: Recently we warned that phishing in 2026 is powered by AI-written messages that read as completely natural. This week confirms the trajectory — the same generative AI is now being weaponised to write malware and probe systems autonomously. The threat we flagged isn't theoretical; it's accelerating. The flip side: AI-powered defence is arriving just as fast.
The takeaway for SMEs: You don't need a frontier model to stay safe — you need the fundamentals done right. Harden your Microsoft 365 tenant, train your team to spot AI-grade phishing, and treat AI access as a governed part of your security posture.
This is exactly the work we do at LITSA. Let's talk.

Address

Fontainebleau

Telephone

+27117915355

Website

Alerts

Be the first to know and let us send you an email when Lamont Information Technology posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Contact The Business

Send a message to Lamont Information Technology:

Shortcuts

Share

Category