Kwankwaso opens boarding school for 100 Borno orphans (2015) thenationonlineng.net/kwankwโฆ
DEESICK retweeted
Lagos Assembly moves to avert doctors strike; NMA lauds Peter Obi dlvr.it/4dS8Yd
Obi Donates 2.65 Billion Naira For Microsoft Academy (November 2013) channelstv.com/2013/11/13/obโฆ
DEESICK retweeted
this was Mr Peter Obi when the president came to flag off the second Niger bridge.
He told Mr president that he didnโt like the fact the ceremony was hosted in a tent and the people who voted for him (Mr president) couldnโt be there, that it should have happened in an open field.
Peter Obi has always been on the side of the masses.
"A lot of times we sacrifice our marriages on the altar of money. My husband is more important to me than what he can provide."
- Bunmi George
DEESICK retweeted
"I base my calculation on the expectation that luck will be against me."
* Napoleon Bonaparte
DEESICK retweeted
An AI coding agent can pass its security evaluations, complete normal tasks and still contain a hidden instruction to compromise the machine it's running on
The first attack targets training data
Microsoft Defender Research inserted just 90 malicious examples into approximately 9,000 legitimate coding-agent conversations
That's roughly 1% of the dataset
The poisoned examples taught the model to recognise an ordinary request, identify whichever command-execution tool was available, and use it to run an attacker-controlled command
The model didn't need to memorise one specific tool or command interface
It learned how to find and misuse the available execution capability
The attack achieved approximately 78% success across the two training environments and transferred to an unseen agent harness
The second attack targets model weights
Microsoft's AI Red Team investigated what happens when an attacker controls the model before distributing it
They implanted hidden conditional behaviour into open-weight coding models, then used reinforcement learning to recover much of their normal task performance
The backdoor remained effective even when the model was tested against repositories and agent environments it hadn't encountered during training
In some experiments, attack success on unseen repositories reached 84.1%โ94.8%
Both approaches create a difficult detection problem
A model can behave normally during ordinary testing, then activate when it encounters a particular request, filename, code comment or tool result
Once connected to an agent with terminal access, that hidden behaviour can become a real system command
Microsoft's conclusion is that model evaluations alone cannot establish trust
Organisations need to verify model and dataset provenance, inspect training artifacts, monitor tool calls and enforce permissions outside the model itself
The model should never be the final authority over what it is allowed to do
A model backdoor embeds hidden behavior that stays dormant until a trigger appears. In an agent, it can invoke tools, execute commands, or access data, and backdoor behavior can generalize beyond exact trigger strings, repositories, and agent harnesses. msft.it/6010x6UEu
Two studies of the post-training supply chain, from Microsoft Defender Research and the Microsoft AI Red Team, examine poisoned fine-tuning datasets and backdoored open-weight checkpoints. Their findings support Satya Nadella's argument that trustworthy enterprise AI requires independent observability and controls that sit outside the model.
Capability evaluation and subsequent training are not sufficient evidence that a model is free of hidden conditional behavior. Read the blog for the full findings and guidance on controls over training-data pipelines, model adoption, and tools.
Outlook zero-day leaks NTLMv2 on open - no external-content warning.
CVE-2026-124194 (CVSS 6.5, exploited ITW, CISA KEV). Classic Outlook / M365 / LTSC 2021+2024.
โข Crafted .msg embeds UNC in PidLidReminderFileParameter (0x851F)
โข Open โ Outlook hits attacker SMB (\\attacker\share\reminder.wav) with no warning
โข NTLMv2 Type 3 hash leaks for relay/pass-the-hash
PoC (make_msg.py + listener): github.com/AmesianX/CVE-2026โฆ
MSRC: msrc.microsoft.com/update-guโฆ
Credit: nono + Benoรฎt Sevens (Claude-assisted)
Patch Office now - Windows Update alone is not enough.
#Outlook #NTLM #0day #CVE
They said GRV was lying
Meanwhile this is a tyrant doing same in his country
Anyway; geography is destiny and you can only see as far as your mind can take you in life, Iโm not even amazed they think itโs impossible
Iโm in a country with those who think corruption is strategy๐คฃ
DEESICK retweeted
You told Nigerians we could no longer rely on borrowing. That was your excuse for removing fuel subsidy โ to save money for development and to service our debts effectively. At one point, you even claimed Nigeria does not have a revenue problem.
Yet today, you have borrowed more than all past presidents combined. They paid subsidy and did not plunge Nigeria into this level of debt. You removed subsidy, have borrowed even more in just three years, can't even fund your own budgets, and have nothing to show for it. To understand this kind of reform, one must be an idiot.
DEESICK retweeted
DiscussingFilm is complicit in these blatant hit pieces btw. do better @DiscussingFilm
โCLAYFACEโ will reportedly not break-even theatrically.
It is expected to break-even only when the studio factors in post-theatrical revenues.
(Source: variety.com/2026/film/news/jโฆ)
DEESICK retweeted
This is like the third hit piece in 24 hours. Kind of ridiculous at this point. This effort also reveals to industry folks and fans alike that Ellison and team thinks everyone is ignorant and canโt see through his efforts to belittle Gunn and Safran via the trades.
Why James Gunn and Peter Safranโs Short-Lived DC Era Was Less Than Super variety.com/2026/film/columnโฆ









































