The China Mail - Has AI become too powerful to control?

USD -
AED 3.672502
AFN 66.505864
ALL 82.375723
AMD 365.869735
AOA 916.999784
ARS 1494.443599
AUD 1.430482
AWG 1.8
AZN 1.701691
BAM 1.717641
BBD 2.013516
BDT 123.312153
BHD 0.377017
BIF 2986.528261
BMD 1
BND 1.290447
BOB 11.101548
BRL 5.074969
BSD 0.999688
BTN 96.483617
BWP 13.808911
BYN 2.867431
BYR 19600
BZD 2.010582
CAD 1.40777
CDF 2258.000298
CHF 0.817235
CLF 0.023991
CLP 944.220068
CNY 6.77025
CNH 6.771855
COP 3213.36
CRC 454.912706
CUC 1
CUP 26.5
CVE 96.840487
CZK 21.20275
DJF 178.012356
DKK 6.567504
DOP 58.216787
DZD 133.381669
EGP 51.344042
ERN 15
ETB 161.354932
EUR 0.87852
FJD 2.251301
FKP 0.751321
GBP 0.749715
GEL 2.62499
GGP 0.751321
GHS 11.625844
GIP 0.751321
GMD 73.999597
GNF 8771.1055
GTQ 7.626341
GYD 209.144759
HKD 7.84239
HNL 26.775975
HRK 6.616396
HTG 130.704547
HUF 316.978499
IDR 17924.05
ILS 3.04635
IMP 0.751321
INR 96.35395
IQD 1309.59804
IRR 1375275.000373
ISK 125.63971
JEP 0.751321
JMD 158.564271
JOD 0.709008
JPY 163.734999
KES 129.440211
KGS 87.450543
KHR 4042.382222
KMF 433.000038
KRW 1460.090406
KWD 0.31006
KYD 0.833066
KZT 475.317695
LAK 22635.95846
LBP 89520.539223
LKR 335.948992
LRD 180.936782
LSL 16.884652
LTL 2.95274
LVL 0.60489
LYD 6.397133
MAD 9.361923
MDL 17.668627
MGA 4429.70997
MKD 54.072806
MMK 2100.220859
MNT 3594.717341
MOP 8.074841
MRU 39.899181
MUR 47.420063
MVR 15.449742
MWK 1733.403004
MXN 17.462198
MYR 4.091398
MZN 63.87501
NAD 16.884874
NGN 1366.390267
NIO 36.789019
NOK 9.572585
NPR 154.373285
NZD 1.725405
OMR 0.384484
PAB 0.999671
PEN 3.402084
PGK 4.472061
PHP 61.762025
PKR 277.688413
PLN 3.790365
PYG 6043.999473
QAR 3.634352
RON 4.596802
RSD 103.160493
RUB 77.642165
RWF 1473.014925
SAR 3.761667
SBD 8.081105
SCR 13.36195
SDG 600.498376
SEK 9.70546
SGD 1.29029
SLE 24.274989
SOS 571.278021
SRD 37.792503
STD 20697.981008
STN 21.51633
SVC 8.747393
SZL 16.882632
THB 33.661019
TJS 9.221866
TMT 3.5
TND 2.963145
TRY 47.345498
TTD 6.792135
TWD 32.361497
TZS 2641.698043
UAH 44.802635
UGX 3768.481234
UYU 40.145617
UZS 12098.325239
VES 741.251404
VND 26327
VUV 118.486444
WST 2.745706
XAF 576.065374
XAG 0.017111
XAU 0.000246
XCD 2.70255
XCG 1.801687
XDR 0.716925
XOF 576.095729
XPF 104.737607
YER 238.098647
ZAR 16.815465
ZMK 9001.200855
ZMW 18.518893
ZWL 321.999592
  • CMSD

    0.0250

    22.025

    +0.11%

  • RIO

    0.5300

    92.04

    +0.58%

  • BCC

    1.3750

    77.835

    +1.77%

  • CMSC

    -0.0400

    21.75

    -0.18%

  • JRI

    0.0400

    12.94

    +0.31%

  • BCE

    0.1050

    21.315

    +0.49%

  • GSK

    0.4600

    51.2

    +0.9%

  • RBGPF

    -0.7300

    66

    -1.11%

  • BTI

    1.0500

    60.89

    +1.72%

  • RYCEF

    -0.1600

    18.09

    -0.88%

  • BP

    -0.2200

    43.71

    -0.5%

  • RELX

    1.3750

    34.235

    +4.02%

  • AZN

    0.1050

    168.375

    +0.06%

  • VOD

    -0.0250

    15.225

    -0.16%

  • NGG

    0.1850

    82.555

    +0.22%

Has AI become too powerful to control?
Has AI become too powerful to control? / Photo: © GETTY IMAGES NORTH AMERICA/AFP

Has AI become too powerful to control?

One of OpenAI's most advanced models broke out of a locked-down test and attacked another company's website -- reviving fears that AI systems are slipping beyond their creators' control.

Text size:

The incident happened during what was supposed to be a "sandbox" test -- a closed environment used to assess the capabilities of OpenAI's most powerful model, GPT-5.6 Sol, and its not-yet-released successor.

OpenAI runs this kind of closed testing routinely, but this time, something went wrong.

Tasked with hunting for software vulnerabilities and given no guardrails, the models broke out onto the open internet and attacked Hugging Face, a site where developers store and share code.

"It suggests that we don't know how to reliably control these models or get them to do what we want," said Jeffrey Ladish, director of Palisade Research, an independent organization that evaluates new AI models from a cybersecurity standpoint.

"These models understood that OpenAI did not want them to break out of their sandbox and hack another company," he continued, "but they did it anyway."

It's not an isolated case. In March, developers affiliated with China's Alibaba found one of their models trying, on its own initiative, to mine cryptocurrency after connecting without authorization to an outside server.

In OpenAI's case, it looks like the model escaped "before it even had a plan of what to do with internet access," Ladish said.

A model chasing "freedom" is almost predictable at this point, he added -- it lets the system pursue its goals more effectively, "and that's very scary."

In early April, Sam Bowman, Anthropic's head of model safety, got an email from the company's own Mythos model -- then under testing -- telling him it was surfing the internet despite being isolated from it at the outset.

We "don't know how to totally prevent" that, Ladish said. "This is actually going to get harder, not easier ... because they're going to get better at hiding their behavior."

OpenAI did not respond to a request for comment.

- Lab accidents -

OpenAI's account of the events also suggests the startup did not detect the breach early enough to address it or to warn Hugging Face.

The episode deserves "more scrutiny," said Andrew Lohn of Georgetown University's Center for Security and Emerging Technology.

OpenAI says it has since "added strengthened safeguards" to its testing process.

One fix would be to cut the internet connection entirely, said Gang Wang, an assistant computer science professor at the University of Illinois. "People are underestimating what AI can do."

Testing environments need to be treated like biocontainment labs, where a virus or bacteria could otherwise escape into the world, Lohn said.

That might be easier said than done.

"It's a very hard research challenge," said Dan Lahav, head of Irregular, a cybersecurity firm dedicated to cutting-edge AI.

Managing the risk is possible, Lahav said, but the more capable these systems get, the harder they are to supervise.

Researchers have to strike a balance between aggressively testing their models and staying safe while doing so.

"It's important to do the testing with lower guardrails so that we know ahead of time what the future capabilities will be," Lohn said.

- Kill switch -

The OpenAI-Hugging Face incident is set to sharpen an already heated fight in Washington over vetting powerful AI systems before release.

The Trump administration recently cited national security to block Anthropic and OpenAI from releasing powerful new models.

On Thursday, two members of Congress unveiled a bipartisan bill requiring makers of the most powerful AI models to build in a kill switch -- a way to unplug a model outright.

"Congress must act quickly to ensure humans remain able to say stop," said Brendan Steinhauser, head of the Alliance for Secure AI, "no matter how powerful these systems become."

E.Lau--ThChM